import os, threading, time, sqlite3, webbrowser, pyautogui, random, cv2
import tkinter as tk
from tkinter import ttk
import tkinter.font as tkFont
from tkinter import filedialog
from tkinter import font
from queue import Queue
from tkinter import Label, Frame, Button
import numpy as np
import pandas as pd
from PIL import Image, ImageOps, ImageTk, ImageDraw, ImageFont, ImageEnhance
from concurrent.futures import ThreadPoolExecutor
from IPython.display import display, HTML
import imageio.v2 as imageio
from collections import deque
from skimage.filters import threshold_otsu
from skimage.exposure import rescale_intensity
from skimage.draw import polygon, line
from skimage.transform import resize
from skimage.morphology import dilation, disk
from skimage.segmentation import find_boundaries
from skimage.util import img_as_ubyte
from scipy.ndimage import binary_fill_holes, label, gaussian_filter
from tkinter import ttk, scrolledtext
from sklearn.model_selection import train_test_split
from xgboost import XGBClassifier
from sklearn.metrics import classification_report, confusion_matrix
[docs]
def restart_gui_app(root):
"""
Restarts the GUI application by destroying the current instance
and launching a fresh one.
"""
try:
# Destroy the current root window
root.destroy()
# Import and launch a new instance of the application
from .gui import gui_app
new_root = tk.Tk() # Create a fresh Tkinter root instance
gui_app()
except Exception as e:
print(f"Error restarting GUI application: {e}")
[docs]
def set_element_size():
screen_width, screen_height = pyautogui.size()
screen_area = screen_width * screen_height
# Calculate sizes based on screen dimensions
btn_size = int((screen_area * 0.002) ** 0.5) # Button size as a fraction of screen area
bar_size = screen_height // 20 # Bar size based on screen height
settings_width = screen_width // 4 # Settings panel width as a fraction of screen width
panel_width = screen_width - settings_width # Panel width as a fraction of screen width
panel_height = screen_height // 6 # Panel height as a fraction of screen height
size_dict = {
'btn_size': btn_size,
'bar_size': bar_size,
'settings_width': settings_width,
'panel_width': panel_width,
'panel_height': panel_height
}
return size_dict
[docs]
def set_dark_style(style, parent_frame=None, containers=None, widgets=None, font_family="OpenSans", font_size=12, bg_color='black', fg_color='white', active_color='blue', inactive_color='dark_gray'):
if active_color == 'teal':
active_color = '#008080'
if inactive_color == 'dark_gray':
inactive_color = '#2B2B2B' # '#333333' #'#050505'
if bg_color == 'black':
bg_color = '#000000'
if fg_color == 'white':
fg_color = '#ffffff'
if active_color == 'blue':
active_color = '#007BFF'
padding = '5 5 5 5'
font_style = tkFont.Font(family=font_family, size=font_size)
if font_family == 'OpenSans':
font_loader = spacrFont(font_name='OpenSans', font_style='Regular', font_size=12)
else:
font_loader = None
style.theme_use('clam')
style.configure('TEntry', padding=padding)
style.configure('TCombobox', padding=padding)
style.configure('Spacr.TEntry', padding=padding)
style.configure('TEntry', padding=padding)
style.configure('Spacr.TEntry', padding=padding)
style.configure('Custom.TLabel', padding=padding)
style.configure('TButton', padding=padding)
style.configure('TFrame', background=bg_color)
style.configure('TPanedwindow', background=bg_color)
if font_loader:
style.configure('TLabel', background=bg_color, foreground=fg_color, font=font_loader.get_font(size=font_size))
else:
style.configure('TLabel', background=bg_color, foreground=fg_color, font=(font_family, font_size))
if parent_frame:
parent_frame.configure(bg=bg_color)
parent_frame.grid_rowconfigure(0, weight=1)
parent_frame.grid_columnconfigure(0, weight=1)
if containers:
for container in containers:
if isinstance(container, ttk.Frame):
container_style = ttk.Style()
container_style.configure(f'{container.winfo_class()}.TFrame', background=bg_color)
container.configure(style=f'{container.winfo_class()}.TFrame')
else:
container.configure(bg=bg_color)
if widgets:
for widget in widgets:
if isinstance(widget, (tk.Label, tk.Button, tk.Frame, ttk.LabelFrame, tk.Canvas)):
widget.configure(bg=bg_color)
if isinstance(widget, (tk.Label, tk.Button)):
if font_loader:
widget.configure(fg=fg_color, font=font_loader.get_font(size=font_size))
else:
widget.configure(fg=fg_color, font=(font_family, font_size))
if isinstance(widget, scrolledtext.ScrolledText):
widget.configure(bg=bg_color, fg=fg_color, insertbackground=fg_color)
if isinstance(widget, tk.OptionMenu):
if font_loader:
widget.configure(bg=bg_color, fg=fg_color, font=font_loader.get_font(size=font_size))
else:
widget.configure(bg=bg_color, fg=fg_color, font=(font_family, font_size))
menu = widget['menu']
if font_loader:
menu.configure(bg=bg_color, fg=fg_color, font=font_loader.get_font(size=font_size))
else:
menu.configure(bg=bg_color, fg=fg_color, font=(font_family, font_size))
return {'font_loader':font_loader, 'font_family': font_family, 'font_size': font_size, 'bg_color': bg_color, 'fg_color': fg_color, 'active_color': active_color, 'inactive_color': inactive_color}
[docs]
class spacrFont:
def __init__(self, font_name, font_style, font_size=12):
"""
Initializes the FontLoader class.
Parameters:
- font_name: str, the name of the font (e.g., 'OpenSans').
- font_style: str, the style of the font (e.g., 'Regular', 'Bold').
- font_size: int, the size of the font (default: 12).
"""
[docs]
self.font_name = font_name
[docs]
self.font_style = font_style
[docs]
self.font_size = font_size
# Determine the path based on the font name and style
[docs]
self.font_path = self.get_font_path(font_name, font_style)
# Register the font with Tkinter
self.load_font()
[docs]
def get_font_path(self, font_name, font_style):
"""
Returns the font path based on the font name and style.
Parameters:
- font_name: str, the name of the font.
- font_style: str, the style of the font.
Returns:
- str, the path to the font file.
"""
base_dir = os.path.dirname(__file__)
if font_name == 'OpenSans':
if font_style == 'Regular':
return os.path.join(base_dir, 'resources/font/open_sans/static/OpenSans-Regular.ttf')
elif font_style == 'Bold':
return os.path.join(base_dir, 'resources/font/open_sans/static/OpenSans-Bold.ttf')
elif font_style == 'Italic':
return os.path.join(base_dir, 'resources/font/open_sans/static/OpenSans-Italic.ttf')
# Add more styles as needed
# Add more fonts as needed
raise ValueError(f"Font '{font_name}' with style '{font_style}' not found.")
[docs]
def load_font(self):
"""
Loads the font into Tkinter.
"""
try:
font.Font(family=self.font_name, size=self.font_size)
except tk.TclError:
# Load the font manually if it's not already loaded
self.tk_font = font.Font(
name=self.font_name,
file=self.font_path,
size=self.font_size
)
[docs]
def get_font(self, size=None):
"""
Returns the font in the specified size.
Parameters:
- size: int, the size of the font (optional).
Returns:
- tkFont.Font object.
"""
if size is None:
size = self.font_size
return font.Font(family=self.font_name, size=size)
[docs]
class spacrContainer(tk.Frame):
def __init__(self, parent, orient=tk.VERTICAL, bg=None, *args, **kwargs):
super().__init__(parent, *args, **kwargs)
[docs]
self.bg = bg if bg else 'lightgrey'
[docs]
self.sash_thickness = 10
self.bind("<Configure>", self.on_configure)
self.grid_rowconfigure(0, weight=1)
self.grid_columnconfigure(0, weight=1)
[docs]
def add(self, widget, stretch='always'):
print(f"Adding widget: {widget} with stretch: {stretch}")
pane = tk.Frame(self, bg=self.bg)
pane.grid_propagate(False)
widget.grid(in_=pane, sticky="nsew") # Use grid for the widget within the pane
self.panes.append((pane, widget))
if len(self.panes) > 1:
self.create_sash()
self.reposition_panes()
[docs]
def create_sash(self):
sash = tk.Frame(self, bg=self.bg, cursor='sb_v_double_arrow' if self.orient == tk.VERTICAL else 'sb_h_double_arrow', height=self.sash_thickness, width=self.sash_thickness)
sash.bind("<Enter>", self.on_enter_sash)
sash.bind("<Leave>", self.on_leave_sash)
sash.bind("<ButtonPress-1>", self.start_resize)
self.sashes.append(sash)
[docs]
def reposition_panes(self):
if not self.panes:
return
total_size = self.winfo_height() if self.orient == tk.VERTICAL else self.winfo_width()
pane_size = total_size // len(self.panes)
print(f"Total size: {total_size}, Pane size: {pane_size}, Number of panes: {len(self.panes)}")
for i, (pane, widget) in enumerate(self.panes):
if self.orient == tk.VERTICAL:
pane.grid(row=i * 2, column=0, sticky="nsew", pady=(0, self.sash_thickness if i < len(self.panes) - 1 else 0))
else:
pane.grid(row=0, column=i * 2, sticky="nsew", padx=(0, self.sash_thickness if i < len(self.panes) - 1 else 0))
for i, sash in enumerate(self.sashes):
if self.orient == tk.VERTICAL:
sash.grid(row=(i * 2) + 1, column=0, sticky="ew")
else:
sash.grid(row=0, column=(i * 2) + 1, sticky="ns")
[docs]
def on_enter_sash(self, event):
event.widget.config(bg='blue')
[docs]
def on_leave_sash(self, event):
event.widget.config(bg=self.bg)
[docs]
def start_resize(self, event):
sash = event.widget
self.start_pos = event.y_root if self.orient == tk.VERTICAL else event.x_root
self.start_size = sash.winfo_y() if self.orient == tk.VERTICAL else sash.winfo_x()
sash.bind("<B1-Motion>", self.perform_resize)
[docs]
class spacrEntry(tk.Frame):
def __init__(self, parent, textvariable=None, outline=False, width=None, *args, **kwargs):
super().__init__(parent, *args, **kwargs)
# Set dark style
style_out = set_dark_style(ttk.Style())
[docs]
self.bg_color = style_out['inactive_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.fg_color = style_out['fg_color']
[docs]
self.outline = outline
[docs]
self.font_family = style_out['font_family']
[docs]
self.font_size = style_out['font_size']
[docs]
self.font_loader = style_out['font_loader']
# Set the background color of the frame
self.configure(bg=style_out['bg_color'])
# Create a canvas for the rounded rectangle background
if width is None:
self.canvas_width = 220 # Adjusted for padding
else:
self.canvas_width = width
[docs]
self.canvas_height = 40 # Adjusted for padding
[docs]
self.canvas = tk.Canvas(self, width=self.canvas_width, height=self.canvas_height, bd=0, highlightthickness=0, relief='ridge', bg=style_out['bg_color'])
self.canvas.pack()
# Create the entry widget
if self.font_loader:
self.entry = tk.Entry(self, textvariable=textvariable, bd=0, highlightthickness=0, fg=self.fg_color, font=self.font_loader.get_font(size=self.font_size), bg=self.bg_color)
else:
self.entry = tk.Entry(self, textvariable=textvariable, bd=0, highlightthickness=0, fg=self.fg_color, font=(self.font_family, self.font_size), bg=self.bg_color)
self.entry.place(relx=0.5, rely=0.5, anchor=tk.CENTER, width=self.canvas_width - 30, height=20) # Centered positioning
# Bind events to change the background color on focus
self.entry.bind("<FocusIn>", self.on_focus_in)
self.entry.bind("<FocusOut>", self.on_focus_out)
self.draw_rounded_rectangle(self.bg_color)
[docs]
def draw_rounded_rectangle(self, color):
radius = 15 # Increased radius for more rounded corners
x0, y0 = 10, 5
x1, y1 = self.canvas_width - 10, self.canvas_height - 5
self.canvas.delete("all")
self.canvas.create_arc((x0, y0, x0 + radius, y0 + radius), start=90, extent=90, fill=color, outline=color)
self.canvas.create_arc((x1 - radius, y0, x1, y0 + radius), start=0, extent=90, fill=color, outline=color)
self.canvas.create_arc((x0, y1 - radius, x0 + radius, y1), start=180, extent=90, fill=color, outline=color)
self.canvas.create_arc((x1 - radius, y1 - radius, x1, y1), start=270, extent=90, fill=color, outline=color)
self.canvas.create_rectangle((x0 + radius / 2, y0, x1 - radius / 2, y1), fill=color, outline=color)
self.canvas.create_rectangle((x0, y0 + radius / 2, x1, y1 - radius / 2), fill=color, outline=color)
[docs]
def on_focus_in(self, event):
self.draw_rounded_rectangle(self.active_color)
self.entry.config(bg=self.active_color)
[docs]
def on_focus_out(self, event):
self.draw_rounded_rectangle(self.bg_color)
self.entry.config(bg=self.bg_color)
[docs]
class spacrCheck(tk.Frame):
def __init__(self, parent, text="", variable=None, *args, **kwargs):
super().__init__(parent, *args, **kwargs)
style_out = set_dark_style(ttk.Style())
[docs]
self.bg_color = style_out['bg_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.fg_color = style_out['fg_color']
[docs]
self.inactive_color = style_out['inactive_color']
[docs]
self.variable = variable
self.configure(bg=self.bg_color)
# Create a canvas for the rounded square background
[docs]
self.canvas_height = 20
[docs]
self.canvas = tk.Canvas(self, width=self.canvas_width, height=self.canvas_height, bd=0, highlightthickness=0, relief='ridge', bg=self.bg_color)
self.canvas.pack()
# Draw the initial rounded square based on the variable's value
self.draw_rounded_square(self.active_color if self.variable.get() else self.inactive_color)
# Bind variable changes to update the checkbox
self.variable.trace_add('write', self.update_check)
# Bind click event to toggle the variable
self.canvas.bind("<Button-1>", self.toggle_variable)
[docs]
def draw_rounded_square(self, color):
radius = 5 # Adjust the radius for more rounded corners
x0, y0 = 2, 2
x1, y1 = 18, 18
self.canvas.delete("all")
self.canvas.create_arc((x0, y0, x0 + radius, y0 + radius), start=90, extent=90, fill=color, outline=self.fg_color)
self.canvas.create_arc((x1 - radius, y0, x1, y0 + radius), start=0, extent=90, fill=color, outline=self.fg_color)
self.canvas.create_arc((x0, y1 - radius, x0 + radius, y1), start=180, extent=90, fill=color, outline=self.fg_color)
self.canvas.create_arc((x1 - radius, y1 - radius, x1, y1), start=270, extent=90, fill=color, outline=self.fg_color)
self.canvas.create_rectangle((x0 + radius / 2, y0, x1 - radius / 2, y1), fill=color, outline=color)
self.canvas.create_rectangle((x0, y0 + radius / 2, x1, y1 - radius / 2), fill=color, outline=color)
self.canvas.create_line(x0 + radius / 2, y0, x1 - radius / 2, y0, fill=self.fg_color)
self.canvas.create_line(x0 + radius / 2, y1, x1 - radius / 2, y1, fill=self.fg_color)
self.canvas.create_line(x0, y0 + radius / 2, x0, y1 - radius / 2, fill=self.fg_color)
self.canvas.create_line(x1, y0 + radius / 2, x1, y1 - radius / 2, fill=self.fg_color)
[docs]
def update_check(self, *args):
self.draw_rounded_square(self.active_color if self.variable.get() else self.inactive_color)
[docs]
def toggle_variable(self, event):
self.variable.set(not self.variable.get())
[docs]
class spacrCombo(tk.Frame):
def __init__(self, parent, textvariable=None, values=None, width=None, *args, **kwargs):
super().__init__(parent, *args, **kwargs)
# Set dark style
style_out = set_dark_style(ttk.Style())
[docs]
self.bg_color = style_out['bg_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.fg_color = style_out['fg_color']
[docs]
self.inactive_color = style_out['inactive_color']
[docs]
self.font_family = style_out['font_family']
[docs]
self.font_size = style_out['font_size']
[docs]
self.font_loader = style_out['font_loader']
[docs]
self.values = values or []
# Create a canvas for the rounded rectangle background
[docs]
self.canvas_width = width if width is not None else 220 # Adjusted for padding
[docs]
self.canvas_height = 40 # Adjusted for padding
[docs]
self.canvas = tk.Canvas(self, width=self.canvas_width, height=self.canvas_height, bd=0, highlightthickness=0, relief='ridge', bg=self.bg_color)
self.canvas.pack()
[docs]
self.var = textvariable if textvariable else tk.StringVar()
[docs]
self.selected_value = self.var.get()
# Create the label to display the selected value
if self.font_loader:
self.label = tk.Label(self, text=self.selected_value, bg=self.inactive_color, fg=self.fg_color, font=self.font_loader.get_font(size=self.font_size))
else:
self.label = tk.Label(self, text=self.selected_value, bg=self.inactive_color, fg=self.fg_color, font=(self.font_family, self.font_size))
self.label.place(relx=0.5, rely=0.5, anchor=tk.CENTER)
# Bind events to open the dropdown menu
self.canvas.bind("<Button-1>", self.on_click)
self.label.bind("<Button-1>", self.on_click)
self.draw_rounded_rectangle(self.inactive_color)
[docs]
def draw_rounded_rectangle(self, color):
radius = 15 # Increased radius for more rounded corners
x0, y0 = 10, 5
x1, y1 = self.canvas_width - 10, self.canvas_height - 5
self.canvas.delete("all")
self.canvas.create_arc((x0, y0, x0 + radius, y0 + radius), start=90, extent=90, fill=color, outline=color)
self.canvas.create_arc((x1 - radius, y0, x1, y0 + radius), start=0, extent=90, fill=color, outline=color)
self.canvas.create_arc((x0, y1 - radius, x0 + radius, y1), start=180, extent=90, fill=color, outline=color)
self.canvas.create_arc((x1 - radius, y1 - radius, x1, y1), start=270, extent=90, fill=color, outline=color)
self.canvas.create_rectangle((x0 + radius / 2, y0, x1 - radius / 2, y1), fill=color, outline=color)
self.canvas.create_rectangle((x0, y0 + radius / 2, x1, y1 - radius / 2), fill=color, outline=color)
self.label.config(bg=color) # Update label background to match rectangle color
[docs]
def on_click(self, event):
if self.dropdown_menu is None:
self.open_dropdown()
else:
self.close_dropdown()
[docs]
def open_dropdown(self):
self.draw_rounded_rectangle(self.active_color)
self.dropdown_menu = tk.Toplevel(self)
self.dropdown_menu.wm_overrideredirect(True)
x, y, width, height = self.winfo_rootx(), self.winfo_rooty(), self.winfo_width(), self.winfo_height()
self.dropdown_menu.geometry(f"{width}x{len(self.values) * 30}+{x}+{y + height}")
for index, value in enumerate(self.values):
display_text = value if value is not None else 'None'
if self.font_loader:
item = tk.Label(self.dropdown_menu, text=display_text, bg=self.inactive_color, fg=self.fg_color, font=self.font_loader.get_font(size=self.font_size), anchor='w')
else:
item = tk.Label(self.dropdown_menu, text=display_text, bg=self.inactive_color, fg=self.fg_color, font=(self.font_family, self.font_size), anchor='w')
item.pack(fill='both')
item.bind("<Button-1>", lambda e, v=value: self.on_select(v))
item.bind("<Enter>", lambda e, w=item: w.config(bg=self.active_color))
item.bind("<Leave>", lambda e, w=item: w.config(bg=self.inactive_color))
[docs]
def close_dropdown(self):
self.draw_rounded_rectangle(self.inactive_color)
if self.dropdown_menu:
self.dropdown_menu.destroy()
self.dropdown_menu = None
[docs]
def on_select(self, value):
display_text = value if value is not None else 'None'
self.var.set(value)
self.label.config(text=display_text)
self.selected_value = value
self.close_dropdown()
[docs]
def set(self, value):
display_text = value if value is not None else 'None'
self.var.set(value)
self.label.config(text=display_text)
self.selected_value = value
[docs]
class spacrProgressBar(ttk.Progressbar):
def __init__(self, parent, label=True, *args, **kwargs):
super().__init__(parent, *args, **kwargs)
# Get the style colors
style_out = set_dark_style(ttk.Style())
[docs]
self.fg_color = style_out['fg_color']
[docs]
self.bg_color = style_out['bg_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.inactive_color = style_out['inactive_color']
[docs]
self.font_size = style_out['font_size']
[docs]
self.font_loader = style_out['font_loader']
# Configure the style for the progress bar
[docs]
self.style = ttk.Style()
# Remove any borders and ensure the active color fills the entire space
self.style.configure(
"spacr.Horizontal.TProgressbar",
troughcolor=self.inactive_color, # Set the trough to bg color
background=self.active_color, # Active part is the active color
borderwidth=0, # Remove border width
pbarrelief="flat", # Flat relief for the progress bar
troughrelief="flat", # Flat relief for the trough
thickness=20, # Set the thickness of the progress bar
darkcolor=self.active_color, # Ensure darkcolor matches the active color
lightcolor=self.active_color, # Ensure lightcolor matches the active color
bordercolor=self.bg_color # Set the border color to the background color to hide it
)
self.configure(style="spacr.Horizontal.TProgressbar")
# Set initial value to 0
self['value'] = 0
# Track whether to show the progress label
# Create the progress label with text wrapping
if self.label:
self.progress_label = tk.Label(
parent,
text="Processing: 0/0",
anchor='w',
justify='left',
bg=self.inactive_color,
fg=self.fg_color,
wraplength=300,
font=self.font_loader.get_font(size=self.font_size)
)
self.progress_label.grid_forget()
# Initialize attributes for time and operation
[docs]
self.operation_type = None
[docs]
self.additional_info = None
[docs]
def set_label_position(self):
if self.label and self.progress_label:
row_info = self.grid_info().get('rowID', 0)
col_info = self.grid_info().get('columnID', 0)
col_span = self.grid_info().get('columnspan', 1)
self.progress_label.grid(row=row_info + 1, column=col_info, columnspan=col_span, pady=5, padx=5, sticky='ew')
[docs]
def update_label(self):
if self.label and self.progress_label:
# Start with the base progress information
label_text = f"Processing: {self['value']}/{self['maximum']}"
# Include the operation type if it exists
if self.operation_type:
label_text += f", {self.operation_type}"
# Handle additional info without adding newlines
if hasattr(self, 'additional_info') and self.additional_info:
# Join all additional info items with a space and ensure they're on the same line
items = self.additional_info.split(", ")
formatted_additional_info = " ".join(items)
# Append the additional info to the label_text, ensuring it's all in one line
label_text += f" {formatted_additional_info.strip()}"
# Update the progress label
self.progress_label.config(text=label_text)
[docs]
class spacrSlider(tk.Frame):
def __init__(self, master=None, length=None, thickness=2, knob_radius=10, position="center", from_=0, to=100, value=None, show_index=False, command=None, **kwargs):
super().__init__(master, **kwargs)
[docs]
self.specified_length = length # Store the specified length, if any
[docs]
self.knob_radius = knob_radius
[docs]
self.thickness = thickness
[docs]
self.knob_position = knob_radius # Start at the beginning of the slider
[docs]
self.slider_line = None
[docs]
self.position = position.lower() # Store the position option
[docs]
self.offset = 0 # Initialize offset
[docs]
self.from_ = from_ # Minimum value of the slider
[docs]
self.to = to # Maximum value of the slider
[docs]
self.value = value if value is not None else from_ # Initial value of the slider
[docs]
self.show_index = show_index # Whether to show the index Entry widget
[docs]
self.command = command # Callback function to handle value changes
# Initialize the style and colors
style_out = set_dark_style(ttk.Style())
[docs]
self.fg_color = style_out['fg_color']
[docs]
self.bg_color = style_out['bg_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.inactive_color = style_out['inactive_color']
# Configure the frame's background color
self.configure(bg=self.bg_color)
# Create a frame for the slider and entry if needed
self.grid_columnconfigure(1, weight=1)
# Entry widget for showing and editing index, if enabled
if self.show_index:
self.index_var = tk.StringVar(value=str(int(self.value)))
self.index_entry = tk.Entry(self, textvariable=self.index_var, width=5, bg=self.bg_color, fg=self.fg_color, insertbackground=self.fg_color)
self.index_entry.grid(row=0, column=0, padx=5)
# Bind the entry to update the slider on change
self.index_entry.bind("<Return>", self.update_slider_from_entry)
# Create the slider canvas
[docs]
self.canvas = tk.Canvas(self, height=knob_radius * 2, bg=self.bg_color, highlightthickness=0)
self.canvas.grid(row=0, column=1, sticky="ew")
# Set initial length to specified length or default value
[docs]
self.length = self.specified_length if self.specified_length is not None else self.canvas.winfo_reqwidth()
# Calculate initial knob position based on the initial value
self.knob_position = self.value_to_position(self.value)
# Bind resize event to dynamically adjust the slider length if no length is specified
self.canvas.bind("<Configure>", self.resize_slider)
# Draw the slider components
self.draw_slider(inactive=True)
# Bind mouse events to the knob and slider
self.canvas.bind("<B1-Motion>", self.move_knob)
self.canvas.bind("<Button-1>", self.activate_knob) # Activate knob on click
self.canvas.bind("<ButtonRelease-1>", self.release_knob) # Trigger command on release
[docs]
def resize_slider(self, event):
if self.specified_length is not None:
self.length = self.specified_length
else:
self.length = int(event.width * 0.9) # 90% of the container width
# Calculate the horizontal offset based on the position
if self.position == "center":
self.offset = (event.width - self.length) // 2
elif self.position == "right":
self.offset = event.width - self.length
else: # position is "left"
self.offset = 0
# Update the knob position after resizing
self.knob_position = self.value_to_position(self.value)
self.draw_slider(inactive=True)
[docs]
def value_to_position(self, value):
if self.to == self.from_:
return self.knob_radius
relative_value = (value - self.from_) / (self.to - self.from_)
return self.knob_radius + relative_value * (self.length - 2 * self.knob_radius)
[docs]
def position_to_value(self, position):
if self.to == self.from_:
return self.from_
relative_position = (position - self.knob_radius) / (self.length - 2 * self.knob_radius)
return self.from_ + relative_position * (self.to - self.from_)
[docs]
def draw_slider(self, inactive=False):
self.canvas.delete("all")
self.slider_line = self.canvas.create_line(
self.offset + self.knob_radius,
self.knob_radius,
self.offset + self.length - self.knob_radius,
self.knob_radius,
fill=self.fg_color,
width=self.thickness
)
knob_color = self.inactive_color if inactive else self.active_color
self.knob = self.canvas.create_oval(
self.offset + self.knob_position - self.knob_radius,
self.knob_radius - self.knob_radius,
self.offset + self.knob_position + self.knob_radius,
self.knob_radius + self.knob_radius,
fill=knob_color,
outline=""
)
[docs]
def move_knob(self, event):
new_position = min(max(event.x - self.offset, self.knob_radius), self.length - self.knob_radius)
self.knob_position = new_position
self.value = self.position_to_value(self.knob_position)
self.canvas.coords(
self.knob,
self.offset + self.knob_position - self.knob_radius,
self.knob_radius - self.knob_radius,
self.offset + self.knob_position + self.knob_radius,
self.knob_radius + self.knob_radius
)
if self.show_index:
self.index_var.set(str(int(self.value)))
[docs]
def activate_knob(self, event):
self.draw_slider(inactive=False)
self.move_knob(event)
[docs]
def release_knob(self, event):
self.draw_slider(inactive=True)
if self.command:
self.command(self.value) # Call the command with the final value when the knob is released
[docs]
def set_to(self, new_to):
self.to = new_to
self.knob_position = self.value_to_position(self.value)
self.draw_slider(inactive=False)
[docs]
def get(self):
return self.value
[docs]
def set(self, value):
"""Set the slider's value and update the knob position."""
self.value = max(self.from_, min(value, self.to)) # Ensure the value is within bounds
self.knob_position = self.value_to_position(self.value)
self.draw_slider(inactive=False)
if self.show_index:
self.index_var.set(str(int(self.value)))
[docs]
def jump_to_click(self, event):
self.activate_knob(event)
[docs]
def update_slider_from_entry(self, event):
"""Update the slider's value from the entry."""
try:
index = int(self.index_var.get())
self.set(index)
if self.command:
self.command(self.value)
except ValueError:
pass
[docs]
class spacrFrame(ttk.Frame):
def __init__(self, container, width=None, *args, bg='black', radius=20, scrollbar=True, textbox=False, **kwargs):
super().__init__(container, *args, **kwargs)
self.configure(style='TFrame')
if width is None:
screen_width = self.winfo_screenwidth()
width = screen_width // 4
# Create the canvas
canvas = tk.Canvas(self, bg=bg, width=width, highlightthickness=0)
self.rounded_rectangle(canvas, 0, 0, width, self.winfo_screenheight(), radius, fill=bg)
# Define scrollbar styles
style_out = set_dark_style(ttk.Style())
[docs]
self.inactive_color = style_out['inactive_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.fg_color = style_out['fg_color'] # Foreground color for text
# Set custom scrollbar style
style = ttk.Style()
spacrScrollbarStyle(style, self.inactive_color, self.active_color)
# Create scrollbar with custom style if scrollbar option is True
if scrollbar:
scrollbar_widget = ttk.Scrollbar(self, orient="vertical", command=canvas.yview, style='Custom.Vertical.TScrollbar')
if textbox:
self.scrollable_frame = tk.Text(canvas, bg=bg, fg=self.fg_color, wrap=tk.WORD)
else:
self.scrollable_frame = ttk.Frame(canvas, style='TFrame')
self.scrollable_frame.bind(
"<Configure>",
lambda e: canvas.configure(scrollregion=canvas.bbox("all"))
)
canvas.create_window((0, 0), window=self.scrollable_frame, anchor="nw")
if scrollbar:
canvas.configure(yscrollcommand=scrollbar_widget.set)
canvas.grid(row=0, column=0, sticky="nsew")
if scrollbar:
scrollbar_widget.grid(row=0, column=1, sticky="ns")
self.grid_rowconfigure(0, weight=1)
self.grid_columnconfigure(0, weight=1)
if scrollbar:
self.grid_columnconfigure(1, weight=0)
_ = set_dark_style(style, containers=[self], widgets=[canvas, self.scrollable_frame])
if scrollbar:
_ = set_dark_style(style, widgets=[scrollbar_widget])
[docs]
def rounded_rectangle(self, canvas, x1, y1, x2, y2, radius=20, **kwargs):
points = [
x1 + radius, y1,
x2 - radius, y1,
x2 - radius, y1,
x2, y1,
x2, y1 + radius,
x2, y2 - radius,
x2, y2 - radius,
x2, y2,
x2 - radius, y2,
x1 + radius, y2,
x1 + radius, y2,
x1, y2,
x1, y2 - radius,
x1, y2 - radius,
x1, y1 + radius,
x1, y1 + radius,
x1, y1
]
return canvas.create_polygon(points, **kwargs, smooth=True)
[docs]
class spacrLabel(tk.Frame):
def __init__(self, parent, text="", font=None, style=None, align="right", height=None, **kwargs):
valid_kwargs = {k: v for k, v in kwargs.items() if k not in ['foreground', 'background', 'font', 'anchor', 'justify', 'wraplength']}
super().__init__(parent, **valid_kwargs)
if height is None:
screen_height = self.winfo_screenheight()
label_height = screen_height // 50
label_width = label_height * 10
else:
label_height = height
label_width = label_height * 10
[docs]
self.style_out = set_dark_style(ttk.Style())
[docs]
self.font_style = self.style_out['font_family']
[docs]
self.font_size = self.style_out['font_size']
[docs]
self.font_family = self.style_out['font_family']
[docs]
self.font_loader = self.style_out['font_loader']
[docs]
self.canvas = tk.Canvas(self, width=label_width, height=label_height, highlightthickness=0, bg=self.style_out['bg_color'])
self.canvas.grid(row=0, column=0, sticky="ew")
if self.style_out['font_family'] != 'OpenSans':
self.font_style = font if font else tkFont.Font(family=self.style_out['font_family'], size=self.style_out['font_size'], weight=tkFont.NORMAL)
if self.align == "center":
anchor_value = tk.CENTER
text_anchor = 'center'
else: # default to right alignment
anchor_value = tk.E
text_anchor = 'e'
if self.style:
ttk_style = ttk.Style()
if self.font_loader:
ttk_style.configure(self.style, font=self.font_loader.get_font(size=self.font_size), background=self.style_out['bg_color'], foreground=self.style_out['fg_color'])
else:
ttk_style.configure(self.style, font=self.font_style, background=self.style_out['bg_color'], foreground=self.style_out['fg_color'])
self.label_text = ttk.Label(self.canvas, text=self.text, style=self.style, anchor=text_anchor)
self.label_text.pack(fill=tk.BOTH, expand=True)
else:
if self.font_loader:
self.label_text = self.canvas.create_text(label_width // 2 if self.align == "center" else label_width - 5,
label_height // 2, text=self.text, fill=self.style_out['fg_color'],
font=self.font_loader.get_font(size=self.font_size), anchor=anchor_value, justify=tk.RIGHT)
else:
self.label_text = self.canvas.create_text(label_width // 2 if self.align == "center" else label_width - 5,
label_height // 2, text=self.text, fill=self.style_out['fg_color'],
font=self.font_style, anchor=anchor_value, justify=tk.RIGHT)
_ = set_dark_style(ttk.Style(), containers=[self], widgets=[self.canvas])
[docs]
def set_text(self, text):
if self.style:
self.label_text.config(text=text)
else:
self.canvas.itemconfig(self.label_text, text=text)
[docs]
class spacrSwitch(ttk.Frame):
def __init__(self, parent, text="", variable=None, command=None, *args, **kwargs):
super().__init__(parent, *args, **kwargs)
[docs]
self.variable = variable if variable else tk.BooleanVar()
[docs]
self.canvas = tk.Canvas(self, width=40, height=20, highlightthickness=0, bd=0)
self.canvas.grid(row=0, column=1, padx=(10, 0))
[docs]
self.switch_bg = self.create_rounded_rectangle(2, 2, 38, 18, radius=9, outline="", fill="#fff")
[docs]
self.switch = self.canvas.create_oval(4, 4, 16, 16, outline="", fill="#800080")
[docs]
self.label = spacrLabel(self, text=self.text)
self.label.grid(row=0, column=0, padx=(0, 10))
self.bind("<Button-1>", self.toggle)
self.canvas.bind("<Button-1>", self.toggle)
self.label.bind("<Button-1>", self.toggle)
self.update_switch()
style = ttk.Style()
_ = set_dark_style(style, containers=[self], widgets=[self.canvas, self.label])
[docs]
def toggle(self, event=None):
self.variable.set(not self.variable.get())
self.animate_switch()
if self.command:
self.command()
[docs]
def update_switch(self):
if self.variable.get():
self.canvas.itemconfig(self.switch, fill="#008080")
self.canvas.coords(self.switch, 24, 4, 36, 16)
else:
self.canvas.itemconfig(self.switch, fill="#800080")
self.canvas.coords(self.switch, 4, 4, 16, 16)
[docs]
def animate_switch(self):
if self.variable.get():
start_x, end_x = 4, 24
final_color = "#008080"
else:
start_x, end_x = 24, 4
final_color = "#800080"
self.animate_movement(start_x, end_x, final_color)
[docs]
def animate_movement(self, start_x, end_x, final_color):
step = 1 if start_x < end_x else -1
for i in range(start_x, end_x, step):
self.canvas.coords(self.switch, i, 4, i + 12, 16)
self.canvas.update()
self.after(10)
self.canvas.itemconfig(self.switch, fill=final_color)
[docs]
def get(self):
return self.variable.get()
[docs]
def set(self, value):
self.variable.set(value)
self.update_switch()
[docs]
def create_rounded_rectangle(self, x1, y1, x2, y2, radius=9, **kwargs):
points = [x1 + radius, y1,
x1 + radius, y1,
x2 - radius, y1,
x2 - radius, y1,
x2, y1,
x2, y1 + radius,
x2, y1 + radius,
x2, y2 - radius,
x2, y2 - radius,
x2, y2,
x2 - radius, y2,
x2 - radius, y2,
x1 + radius, y2,
x1 + radius, y2,
x1, y2,
x1, y2 - radius,
x1, y2 - radius,
x1, y1 + radius,
x1, y1 + radius,
x1, y1]
return self.canvas.create_polygon(points, **kwargs, smooth=True)
[docs]
class ModifyMaskApp:
def __init__(self, root, folder_path, scale_factor):
[docs]
self.folder_path = folder_path
[docs]
self.scale_factor = scale_factor
[docs]
self.image_filenames = sorted([f for f in os.listdir(folder_path) if f.endswith(('.png', '.jpg', '.jpeg', '.tif', '.tiff'))])
[docs]
self.masks_folder = os.path.join(folder_path, 'masks')
[docs]
self.current_image_index = 0
self.initialize_flags()
[docs]
self.canvas_width = self.root.winfo_screenheight() -100
[docs]
self.canvas_height = self.root.winfo_screenheight() -100
self.root.configure(bg='black')
self.setup_navigation_toolbar()
self.setup_mode_toolbar()
self.setup_function_toolbar()
self.setup_zoom_toolbar()
self.setup_canvas()
self.load_first_image()
####################################################################################################
# Helper functions#
####################################################################################################
[docs]
def update_display(self):
if self.zoom_active:
self.display_zoomed_image()
else:
self.display_image()
[docs]
def update_original_mask_from_zoom(self):
y0, y1, x0, x1 = self.zoom_y0, self.zoom_y1, self.zoom_x0, self.zoom_x1
zoomed_mask_resized = resize(self.zoom_mask, (y1 - y0, x1 - x0), order=0, preserve_range=True).astype(np.uint8)
self.mask[y0:y1, x0:x1] = zoomed_mask_resized
[docs]
def update_original_mask(self, zoomed_mask, x0, x1, y0, y1):
actual_mask_region = self.mask[y0:y1, x0:x1]
target_shape = actual_mask_region.shape
resized_mask = resize(zoomed_mask, target_shape, order=0, preserve_range=True).astype(np.uint8)
if resized_mask.shape != actual_mask_region.shape:
raise ValueError(f"Shape mismatch: resized_mask {resized_mask.shape}, actual_mask_region {actual_mask_region.shape}")
self.mask[y0:y1, x0:x1] = np.maximum(actual_mask_region, resized_mask)
self.mask = self.mask.copy()
self.mask[y0:y1, x0:x1] = np.maximum(self.mask[y0:y1, x0:x1], resized_mask)
self.mask = self.mask.copy()
[docs]
def get_scaling_factors(self, img_width, img_height, canvas_width, canvas_height):
x_scale = img_width / canvas_width
y_scale = img_height / canvas_height
return x_scale, y_scale
[docs]
def canvas_to_image(self, x_canvas, y_canvas):
x_scale, y_scale = self.get_scaling_factors(
self.image.shape[1], self.image.shape[0],
self.canvas_width, self.canvas_height
)
x_image = int(x_canvas * x_scale)
y_image = int(y_canvas * y_scale)
return x_image, y_image
[docs]
def apply_zoom_on_enter(self, event):
if self.zoom_active and self.zoom_rectangle_start is not None:
self.set_zoom_rectangle_end(event)
[docs]
def normalize_image(self, image, lower_quantile, upper_quantile):
lower_bound = np.percentile(image, lower_quantile)
upper_bound = np.percentile(image, upper_quantile)
normalized = np.clip(image, lower_bound, upper_bound)
normalized = (normalized - lower_bound) / (upper_bound - lower_bound)
max_value = np.iinfo(image.dtype).max
normalized = (normalized * max_value).astype(image.dtype)
return normalized
[docs]
def resize_arrays(self, img, mask):
original_dtype = img.dtype
scaled_height = int(img.shape[0] * self.scale_factor)
scaled_width = int(img.shape[1] * self.scale_factor)
scaled_img = resize(img, (scaled_height, scaled_width), anti_aliasing=True, preserve_range=True)
scaled_mask = resize(mask, (scaled_height, scaled_width), order=0, anti_aliasing=False, preserve_range=True)
stretched_img = resize(scaled_img, (self.canvas_height, self.canvas_width), anti_aliasing=True, preserve_range=True)
stretched_mask = resize(scaled_mask, (self.canvas_height, self.canvas_width), order=0, anti_aliasing=False, preserve_range=True)
return stretched_img.astype(original_dtype), stretched_mask.astype(original_dtype)
####################################################################################################
#Initiate canvas elements#
####################################################################################################
[docs]
def load_first_image(self):
self.image, self.mask = self.load_image_and_mask(self.current_image_index)
self.original_size = self.image.shape
self.image, self.mask = self.resize_arrays(self.image, self.mask)
self.display_image()
[docs]
def setup_canvas(self):
self.canvas = tk.Canvas(self.root, width=self.canvas_width, height=self.canvas_height, bg='black')
self.canvas.pack()
self.canvas.bind("<Motion>", self.update_mouse_info)
[docs]
def initialize_flags(self):
self.zoom_rectangle_start = None
self.zoom_rectangle_end = None
self.zoom_rectangle_id = None
self.zoom_x0 = None
self.zoom_y0 = None
self.zoom_x1 = None
self.zoom_y1 = None
self.zoom_mask = None
self.zoom_image = None
self.zoom_image_orig = None
self.zoom_scale = 1
self.drawing = False
self.zoom_active = False
self.magic_wand_active = False
self.brush_active = False
self.dividing_line_active = False
self.dividing_line_coords = []
self.current_dividing_line = None
self.lower_quantile = tk.StringVar(value="1.0")
self.upper_quantile = tk.StringVar(value="99.9")
self.magic_wand_tolerance = tk.StringVar(value="1000")
[docs]
def update_mouse_info(self, event):
x, y = event.x, event.y
intensity = "N/A"
mask_value = "N/A"
pixel_count = "N/A"
if self.zoom_active:
if 0 <= x < self.canvas_width and 0 <= y < self.canvas_height:
intensity = self.zoom_image_orig[y, x] if self.zoom_image_orig is not None else "N/A"
mask_value = self.zoom_mask[y, x] if self.zoom_mask is not None else "N/A"
else:
if 0 <= x < self.image.shape[1] and 0 <= y < self.image.shape[0]:
intensity = self.image[y, x]
mask_value = self.mask[y, x]
if mask_value != "N/A" and mask_value != 0:
pixel_count = np.sum(self.mask == mask_value)
self.intensity_label.config(text=f"Intensity: {intensity}")
self.mask_value_label.config(text=f"Mask: {mask_value}, Area: {pixel_count}")
self.mask_value_label.config(text=f"Mask: {mask_value}")
if mask_value != "N/A" and mask_value != 0:
self.pixel_count_label.config(text=f"Area: {pixel_count}")
else:
self.pixel_count_label.config(text="Area: N/A")
[docs]
def load_image_and_mask(self, index):
# Load the image
image_path = os.path.join(self.folder_path, self.image_filenames[index])
image = imageio.imread(image_path)
print(f"Original Image shape: {image.shape}, dtype: {image.dtype}")
# Handle multi-channel or transparency issues
if image.ndim == 3:
if image.shape[2] == 4: # If the image has an alpha channel (RGBA)
image = image[..., :3] # Remove the alpha channel
# Convert RGB to grayscale using weighted average
image = np.dot(image[..., :3], [0.2989, 0.5870, 0.1140]).astype(np.uint8)
print(f"Converted to grayscale: {image.shape}")
# Ensure the shape is (height, width) without extra channel
if image.ndim == 3 and image.shape[2] == 1:
image = np.squeeze(image, axis=-1)
if image.dtype != np.uint16:
# Scale the image to fit the 16-bit range (0–65535)
image = (image / image.max() * 65535).astype(np.uint16)
# eventually remove this images should not have to be 16 bit look into downstream function (non 16bit images are jsut black)
# Load the corresponding mask
mask_path = os.path.join(self.masks_folder, self.image_filenames[index])
if os.path.exists(mask_path):
print(f'Loading mask: {mask_path} for image: {image_path}')
mask = imageio.imread(mask_path)
# Ensure mask is uint8
if mask.dtype != np.uint8:
mask = (mask / mask.max() * 255).astype(np.uint8)
else:
# Create a new mask with the same size as the image
mask = np.zeros(image.shape[:2], dtype=np.uint8)
print(f'Loaded new mask for image: {image_path}')
return image, mask
####################################################################################################
# Image Display functions#
####################################################################################################
[docs]
def display_image(self):
if self.zoom_rectangle_id is not None:
self.canvas.delete(self.zoom_rectangle_id)
self.zoom_rectangle_id = None
lower_quantile = float(self.lower_quantile.get()) if self.lower_quantile.get() else 1.0
upper_quantile = float(self.upper_quantile.get()) if self.upper_quantile.get() else 99.9
normalized = self.normalize_image(self.image, lower_quantile, upper_quantile)
combined = self.overlay_mask_on_image(normalized, self.mask)
self.tk_image = ImageTk.PhotoImage(image=Image.fromarray(combined))
self.canvas.create_image(0, 0, anchor='nw', image=self.tk_image)
[docs]
def display_zoomed_image(self):
if self.zoom_rectangle_start and self.zoom_rectangle_end:
# Convert canvas coordinates to image coordinates
x0, y0 = self.canvas_to_image(*self.zoom_rectangle_start)
x1, y1 = self.canvas_to_image(*self.zoom_rectangle_end)
x0, x1 = min(x0, x1), max(x0, x1)
y0, y1 = min(y0, y1), max(y0, y1)
self.zoom_x0 = x0
self.zoom_y0 = y0
self.zoom_x1 = x1
self.zoom_y1 = y1
# Normalize the entire image
lower_quantile = float(self.lower_quantile.get()) if self.lower_quantile.get() else 1.0
upper_quantile = float(self.upper_quantile.get()) if self.upper_quantile.get() else 99.9
normalized_image = self.normalize_image(self.image, lower_quantile, upper_quantile)
# Extract the zoomed portion of the normalized image and mask
self.zoom_image = normalized_image[y0:y1, x0:x1]
self.zoom_image_orig = self.image[y0:y1, x0:x1]
self.zoom_mask = self.mask[y0:y1, x0:x1]
original_mask_area = self.mask.shape[0] * self.mask.shape[1]
zoom_mask_area = self.zoom_mask.shape[0] * self.zoom_mask.shape[1]
if original_mask_area > 0:
self.zoom_scale = original_mask_area/zoom_mask_area
# Resize the zoomed image and mask to fit the canvas
canvas_height = self.canvas.winfo_height()
canvas_width = self.canvas.winfo_width()
if self.zoom_image.size > 0 and canvas_height > 0 and canvas_width > 0:
self.zoom_image = resize(self.zoom_image, (canvas_height, canvas_width), preserve_range=True).astype(self.zoom_image.dtype)
self.zoom_image_orig = resize(self.zoom_image_orig, (canvas_height, canvas_width), preserve_range=True).astype(self.zoom_image_orig.dtype)
#self.zoom_mask = resize(self.zoom_mask, (canvas_height, canvas_width), preserve_range=True).astype(np.uint8)
self.zoom_mask = resize(self.zoom_mask, (canvas_height, canvas_width), order=0, preserve_range=True).astype(np.uint8)
combined = self.overlay_mask_on_image(self.zoom_image, self.zoom_mask)
self.tk_image = ImageTk.PhotoImage(image=Image.fromarray(combined))
self.canvas.create_image(0, 0, anchor='nw', image=self.tk_image)
[docs]
def overlay_mask_on_image(self, image, mask, alpha=0.5):
if len(image.shape) == 2:
image = np.stack((image,) * 3, axis=-1)
mask = mask.astype(np.int32)
max_label = np.max(mask)
np.random.seed(0)
colors = np.random.randint(0, 255, size=(max_label + 1, 3), dtype=np.uint8)
colors[0] = [0, 0, 0] # background color
colored_mask = colors[mask]
image_8bit = (image / 256).astype(np.uint8)
# Blend the mask and the image with transparency
combined_image = np.where(mask[..., None] > 0,
np.clip(image_8bit * (1 - alpha) + colored_mask * alpha, 0, 255),
image_8bit)
# Convert the final image back to uint8
combined_image = combined_image.astype(np.uint8)
return combined_image
####################################################################################################
# Navigation functions#
####################################################################################################
[docs]
def previous_image(self):
if self.current_image_index > 0:
self.current_image_index -= 1
self.initialize_flags()
self.image, self.mask = self.load_image_and_mask(self.current_image_index)
self.original_size = self.image.shape
self.image, self.mask = self.resize_arrays(self.image, self.mask)
self.display_image()
[docs]
def next_image(self):
if self.current_image_index < len(self.image_filenames) - 1:
self.current_image_index += 1
self.initialize_flags()
self.image, self.mask = self.load_image_and_mask(self.current_image_index)
self.original_size = self.image.shape
self.image, self.mask = self.resize_arrays(self.image, self.mask)
self.display_image()
[docs]
def save_mask(self):
if self.current_image_index < len(self.image_filenames):
original_size = self.original_size
if self.mask.shape != original_size:
resized_mask = resize(self.mask, original_size, order=0, preserve_range=True).astype(np.uint16)
else:
resized_mask = self.mask
resized_mask, _ = label(resized_mask > 0)
save_folder = os.path.join(self.folder_path, 'masks')
if not os.path.exists(save_folder):
os.makedirs(save_folder)
image_filename = os.path.splitext(self.image_filenames[self.current_image_index])[0] + '.tif'
save_path = os.path.join(save_folder, image_filename)
print(f"Saving mask to: {save_path}") # Debug print
imageio.imwrite(save_path, resized_mask)
####################################################################################################
# Zoom Functions #
####################################################################################################
[docs]
def set_zoom_rectangle_start(self, event):
if self.zoom_active:
self.zoom_rectangle_start = (event.x, event.y)
[docs]
def set_zoom_rectangle_end(self, event):
if self.zoom_active:
self.zoom_rectangle_end = (event.x, event.y)
if self.zoom_rectangle_id is not None:
self.canvas.delete(self.zoom_rectangle_id)
self.zoom_rectangle_id = None
self.display_zoomed_image()
self.canvas.unbind("<Motion>")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<Button-3>")
self.canvas.bind("<Motion>", self.update_mouse_info)
[docs]
def update_zoom_box(self, event):
if self.zoom_active and self.zoom_rectangle_start is not None:
if self.zoom_rectangle_id is not None:
self.canvas.delete(self.zoom_rectangle_id)
# Assuming event.x and event.y are already in image coordinates
self.zoom_rectangle_end = (event.x, event.y)
x0, y0 = self.zoom_rectangle_start
x1, y1 = self.zoom_rectangle_end
self.zoom_rectangle_id = self.canvas.create_rectangle(x0, y0, x1, y1, outline="red", width=2)
####################################################################################################
# Mode activation#
####################################################################################################
[docs]
def toggle_zoom_mode(self):
if not self.zoom_active:
self.brush_btn.config(text="Brush")
self.canvas.unbind("<B1-Motion>")
self.canvas.unbind("<B3-Motion>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<ButtonRelease-3>")
self.zoom_active = True
self.drawing = False
self.magic_wand_active = False
self.erase_active = False
self.brush_active = False
self.dividing_line_active = False
self.draw_btn.config(text="Draw")
self.erase_btn.config(text="Erase")
self.magic_wand_btn.config(text="Magic Wand")
self.zoom_btn.config(text="Zoom ON")
self.dividing_line_btn.config(text="Dividing Line")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<Button-3>")
self.canvas.unbind("<Motion>")
self.canvas.bind("<Button-1>", self.set_zoom_rectangle_start)
self.canvas.bind("<Button-3>", self.set_zoom_rectangle_end)
self.canvas.bind("<Motion>", self.update_zoom_box)
else:
self.zoom_active = False
self.zoom_btn.config(text="Zoom")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<Button-3>")
self.canvas.unbind("<Motion>")
self.zoom_rectangle_start = self.zoom_rectangle_end = None
self.zoom_rectangle_id = None
self.display_image()
self.canvas.bind("<Motion>", self.update_mouse_info)
self.zoom_rectangle_start = None
self.zoom_rectangle_end = None
self.zoom_rectangle_id = None
self.zoom_x0 = None
self.zoom_y0 = None
self.zoom_x1 = None
self.zoom_y1 = None
self.zoom_mask = None
self.zoom_image = None
self.zoom_image_orig = None
[docs]
def toggle_brush_mode(self):
self.brush_active = not self.brush_active
if self.brush_active:
self.drawing = False
self.magic_wand_active = False
self.erase_active = False
self.brush_btn.config(text="Brush ON")
self.draw_btn.config(text="Draw")
self.erase_btn.config(text="Erase")
self.magic_wand_btn.config(text="Magic Wand")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<Button-3>")
self.canvas.unbind("<Motion>")
self.canvas.bind("<B1-Motion>", self.apply_brush) # Left click and drag to apply brush
self.canvas.bind("<B3-Motion>", self.erase_brush) # Right click and drag to erase with brush
self.canvas.bind("<ButtonRelease-1>", self.apply_brush_release) # Left button release
self.canvas.bind("<ButtonRelease-3>", self.erase_brush_release) # Right button release
else:
self.brush_active = False
self.brush_btn.config(text="Brush")
self.canvas.unbind("<B1-Motion>")
self.canvas.unbind("<B3-Motion>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<ButtonRelease-3>")
[docs]
def image_to_canvas(self, x_image, y_image):
x_scale, y_scale = self.get_scaling_factors(
self.image.shape[1], self.image.shape[0],
self.canvas_width, self.canvas_height
)
x_canvas = int(x_image / x_scale)
y_canvas = int(y_image / y_scale)
return x_canvas, y_canvas
[docs]
def toggle_dividing_line_mode(self):
self.dividing_line_active = not self.dividing_line_active
if self.dividing_line_active:
self.drawing = False
self.magic_wand_active = False
self.erase_active = False
self.brush_active = False
self.draw_btn.config(text="Draw")
self.erase_btn.config(text="Erase")
self.magic_wand_btn.config(text="Magic Wand")
self.brush_btn.config(text="Brush")
self.dividing_line_btn.config(text="Dividing Line ON")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<Motion>")
self.canvas.bind("<Button-1>", self.start_dividing_line)
self.canvas.bind("<ButtonRelease-1>", self.finish_dividing_line)
self.canvas.bind("<Motion>", self.update_dividing_line_preview)
else:
print("Dividing Line Mode: OFF")
self.dividing_line_active = False
self.dividing_line_btn.config(text="Dividing Line")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<Motion>")
self.display_image()
[docs]
def start_dividing_line(self, event):
if self.dividing_line_active:
self.dividing_line_coords = [(event.x, event.y)]
self.current_dividing_line = self.canvas.create_line(event.x, event.y, event.x, event.y, fill="red", width=2)
[docs]
def finish_dividing_line(self, event):
if self.dividing_line_active:
self.dividing_line_coords.append((event.x, event.y))
if self.zoom_active:
self.dividing_line_coords = [self.canvas_to_image(x, y) for x, y in self.dividing_line_coords]
self.apply_dividing_line()
self.canvas.delete(self.current_dividing_line)
self.current_dividing_line = None
[docs]
def update_dividing_line_preview(self, event):
if self.dividing_line_active and self.dividing_line_coords:
x, y = event.x, event.y
if self.zoom_active:
x, y = self.canvas_to_image(x, y)
self.dividing_line_coords.append((x, y))
canvas_coords = [(self.image_to_canvas(*pt) if self.zoom_active else pt) for pt in self.dividing_line_coords]
flat_canvas_coords = [coord for pt in canvas_coords for coord in pt]
self.canvas.coords(self.current_dividing_line, *flat_canvas_coords)
[docs]
def apply_dividing_line(self):
if self.dividing_line_coords:
coords = self.dividing_line_coords
if self.zoom_active:
coords = [self.canvas_to_image(x, y) for x, y in coords]
rr, cc = [], []
for (x0, y0), (x1, y1) in zip(coords[:-1], coords[1:]):
line_rr, line_cc = line(y0, x0, y1, x1)
rr.extend(line_rr)
cc.extend(line_cc)
rr, cc = np.array(rr), np.array(cc)
mask_copy = self.mask.copy()
if self.zoom_active:
# Update the zoomed mask
self.zoom_mask[rr, cc] = 0
# Reflect changes to the original mask
y0, y1, x0, x1 = self.zoom_y0, self.zoom_y1, self.zoom_x0, self.zoom_x1
zoomed_mask_resized_back = resize(self.zoom_mask, (y1 - y0, x1 - x0), order=0, preserve_range=True).astype(np.uint8)
self.mask[y0:y1, x0:x1] = zoomed_mask_resized_back
else:
# Directly update the original mask
mask_copy[rr, cc] = 0
self.mask = mask_copy
labeled_mask, num_labels = label(self.mask > 0)
self.mask = labeled_mask
self.update_display()
self.dividing_line_coords = []
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<Motion>")
self.dividing_line_active = False
self.dividing_line_btn.config(text="Dividing Line")
[docs]
def toggle_draw_mode(self):
self.drawing = not self.drawing
if self.drawing:
self.brush_btn.config(text="Brush")
self.canvas.unbind("<B1-Motion>")
self.canvas.unbind("<B3-Motion>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<ButtonRelease-3>")
self.magic_wand_active = False
self.erase_active = False
self.brush_active = False
self.draw_btn.config(text="Draw ON")
self.magic_wand_btn.config(text="Magic Wand")
self.erase_btn.config(text="Erase")
self.draw_coordinates = []
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<Motion>")
self.canvas.bind("<B1-Motion>", self.draw)
self.canvas.bind("<ButtonRelease-1>", self.finish_drawing)
else:
self.drawing = False
self.draw_btn.config(text="Draw")
self.canvas.unbind("<B1-Motion>")
self.canvas.unbind("<ButtonRelease-1>")
[docs]
def toggle_magic_wand_mode(self):
self.magic_wand_active = not self.magic_wand_active
if self.magic_wand_active:
self.brush_btn.config(text="Brush")
self.canvas.unbind("<B1-Motion>")
self.canvas.unbind("<B3-Motion>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<ButtonRelease-3>")
self.drawing = False
self.erase_active = False
self.brush_active = False
self.draw_btn.config(text="Draw")
self.erase_btn.config(text="Erase")
self.magic_wand_btn.config(text="Magic Wand ON")
self.canvas.bind("<Button-1>", self.use_magic_wand)
self.canvas.bind("<Button-3>", self.use_magic_wand)
else:
self.magic_wand_btn.config(text="Magic Wand")
self.canvas.unbind("<Button-1>")
self.canvas.unbind("<Button-3>")
[docs]
def toggle_erase_mode(self):
self.erase_active = not self.erase_active
if self.erase_active:
self.brush_btn.config(text="Brush")
self.canvas.unbind("<B1-Motion>")
self.canvas.unbind("<B3-Motion>")
self.canvas.unbind("<ButtonRelease-1>")
self.canvas.unbind("<ButtonRelease-3>")
self.erase_btn.config(text="Erase ON")
self.canvas.bind("<Button-1>", self.erase_object)
self.drawing = False
self.magic_wand_active = False
self.brush_active = False
self.draw_btn.config(text="Draw")
self.magic_wand_btn.config(text="Magic Wand")
else:
self.erase_active = False
self.erase_btn.config(text="Erase")
self.canvas.unbind("<Button-1>")
####################################################################################################
# Mode functions#
####################################################################################################
[docs]
def apply_brush_release(self, event):
if hasattr(self, 'brush_path'):
for x, y, brush_size in self.brush_path:
img_x, img_y = (x, y) if self.zoom_active else self.canvas_to_image(x, y)
x0 = max(img_x - brush_size // 2, 0)
y0 = max(img_y - brush_size // 2, 0)
x1 = min(img_x + brush_size // 2, self.zoom_mask.shape[1] if self.zoom_active else self.mask.shape[1])
y1 = min(img_y + brush_size // 2, self.zoom_mask.shape[0] if self.zoom_active else self.mask.shape[0])
if self.zoom_active:
self.zoom_mask[y0:y1, x0:x1] = 255
self.update_original_mask_from_zoom()
else:
self.mask[y0:y1, x0:x1] = 255
del self.brush_path
self.canvas.delete("temp_line")
self.update_display()
[docs]
def erase_brush_release(self, event):
if hasattr(self, 'erase_path'):
for x, y, brush_size in self.erase_path:
img_x, img_y = (x, y) if self.zoom_active else self.canvas_to_image(x, y)
x0 = max(img_x - brush_size // 2, 0)
y0 = max(img_y - brush_size // 2, 0)
x1 = min(img_x + brush_size // 2, self.zoom_mask.shape[1] if self.zoom_active else self.mask.shape[1])
y1 = min(img_y + brush_size // 2, self.zoom_mask.shape[0] if self.zoom_active else self.mask.shape[0])
if self.zoom_active:
self.zoom_mask[y0:y1, x0:x1] = 0
self.update_original_mask_from_zoom()
else:
self.mask[y0:y1, x0:x1] = 0
del self.erase_path
self.canvas.delete("temp_line")
self.update_display()
[docs]
def apply_brush(self, event):
brush_size = int(self.brush_size_entry.get())
x, y = event.x, event.y
if not hasattr(self, 'brush_path'):
self.brush_path = []
self.last_brush_coord = (x, y)
if self.last_brush_coord:
last_x, last_y = self.last_brush_coord
rr, cc = line(last_y, last_x, y, x)
for ry, rx in zip(rr, cc):
self.brush_path.append((rx, ry, brush_size))
self.canvas.create_line(self.last_brush_coord[0], self.last_brush_coord[1], x, y, width=brush_size, fill="blue", tag="temp_line")
self.last_brush_coord = (x, y)
[docs]
def erase_brush(self, event):
brush_size = int(self.brush_size_entry.get())
x, y = event.x, event.y
if not hasattr(self, 'erase_path'):
self.erase_path = []
self.last_erase_coord = (x, y)
if self.last_erase_coord:
last_x, last_y = self.last_erase_coord
rr, cc = line(last_y, last_x, y, x)
for ry, rx in zip(rr, cc):
self.erase_path.append((rx, ry, brush_size))
self.canvas.create_line(self.last_erase_coord[0], self.last_erase_coord[1], x, y, width=brush_size, fill="white", tag="temp_line")
self.last_erase_coord = (x, y)
[docs]
def erase_object(self, event):
x, y = event.x, event.y
if self.zoom_active:
canvas_x, canvas_y = x, y
zoomed_x = int(canvas_x * (self.zoom_image.shape[1] / self.canvas_width))
zoomed_y = int(canvas_y * (self.zoom_image.shape[0] / self.canvas_height))
orig_x = int(zoomed_x * ((self.zoom_x1 - self.zoom_x0) / self.canvas_width) + self.zoom_x0)
orig_y = int(zoomed_y * ((self.zoom_y1 - self.zoom_y0) / self.canvas_height) + self.zoom_y0)
if orig_x < 0 or orig_y < 0 or orig_x >= self.image.shape[1] or orig_y >= self.image.shape[0]:
print("Point is out of bounds in the original image.")
return
else:
orig_x, orig_y = x, y
label_to_remove = self.mask[orig_y, orig_x]
if label_to_remove > 0:
self.mask[self.mask == label_to_remove] = 0
self.update_display()
[docs]
def use_magic_wand(self, event):
x, y = event.x, event.y
tolerance = int(self.magic_wand_tolerance.get())
maximum = int(self.max_pixels_entry.get())
action = 'add' if event.num == 1 else 'erase'
if self.zoom_active:
self.magic_wand_zoomed((x, y), tolerance, action)
else:
self.magic_wand_normal((x, y), tolerance, action)
[docs]
def apply_magic_wand(self, image, mask, seed_point, tolerance, maximum, action='add'):
x, y = seed_point
initial_value = image[y, x].astype(np.float32)
visited = np.zeros_like(image, dtype=bool)
queue = deque([(x, y)])
added_pixels = 0
while queue and added_pixels < maximum:
cx, cy = queue.popleft()
if visited[cy, cx]:
continue
visited[cy, cx] = True
current_value = image[cy, cx].astype(np.float32)
if np.linalg.norm(abs(current_value - initial_value)) <= tolerance:
if mask[cy, cx] == 0:
added_pixels += 1
mask[cy, cx] = 255 if action == 'add' else 0
if added_pixels >= maximum:
break
for dx, dy in [(-1, 0), (1, 0), (0, -1), (0, 1)]:
nx, ny = cx + dx, cy + dy
if 0 <= nx < image.shape[1] and 0 <= ny < image.shape[0] and not visited[ny, nx]:
queue.append((nx, ny))
return mask
[docs]
def magic_wand_normal(self, seed_point, tolerance, action):
try:
maximum = int(self.max_pixels_entry.get())
except ValueError:
print("Invalid maximum value; using default of 1000")
maximum = 1000
self.mask = self.apply_magic_wand(self.image, self.mask, seed_point, tolerance, maximum, action)
self.display_image()
[docs]
def magic_wand_zoomed(self, seed_point, tolerance, action):
if self.zoom_image_orig is None or self.zoom_mask is None:
print("Zoomed image or mask not initialized")
return
try:
maximum = int(self.max_pixels_entry.get())
maximum = maximum * self.zoom_scale
except ValueError:
print("Invalid maximum value; using default of 1000")
maximum = 1000
canvas_x, canvas_y = seed_point
if canvas_x < 0 or canvas_y < 0 or canvas_x >= self.zoom_image_orig.shape[1] or canvas_y >= self.zoom_image_orig.shape[0]:
print("Selected point is out of bounds in the zoomed image.")
return
self.zoom_mask = self.apply_magic_wand(self.zoom_image_orig, self.zoom_mask, (canvas_x, canvas_y), tolerance, maximum, action)
y0, y1, x0, x1 = self.zoom_y0, self.zoom_y1, self.zoom_x0, self.zoom_x1
zoomed_mask_resized_back = resize(self.zoom_mask, (y1 - y0, x1 - x0), order=0, preserve_range=True).astype(np.uint8)
if action == 'erase':
self.mask[y0:y1, x0:x1] = np.where(zoomed_mask_resized_back == 0, 0, self.mask[y0:y1, x0:x1])
else:
self.mask[y0:y1, x0:x1] = np.where(zoomed_mask_resized_back > 0, zoomed_mask_resized_back, self.mask[y0:y1, x0:x1])
self.update_display()
[docs]
def draw(self, event):
if self.drawing:
x, y = event.x, event.y
if self.draw_coordinates:
last_x, last_y = self.draw_coordinates[-1]
self.current_line = self.canvas.create_line(last_x, last_y, x, y, fill="yellow", width=3)
self.draw_coordinates.append((x, y))
[docs]
def draw_on_zoomed_mask(self, draw_coordinates):
canvas_height = self.canvas.winfo_height()
canvas_width = self.canvas.winfo_width()
zoomed_mask = np.zeros((canvas_height, canvas_width), dtype=np.uint8)
rr, cc = polygon(np.array(draw_coordinates)[:, 1], np.array(draw_coordinates)[:, 0], shape=zoomed_mask.shape)
zoomed_mask[rr, cc] = 255
return zoomed_mask
[docs]
def finish_drawing(self, event):
if len(self.draw_coordinates) > 2:
self.draw_coordinates.append(self.draw_coordinates[0])
if self.zoom_active:
x0, x1, y0, y1 = self.zoom_x0, self.zoom_x1, self.zoom_y0, self.zoom_y1
zoomed_mask = self.draw_on_zoomed_mask(self.draw_coordinates)
self.update_original_mask(zoomed_mask, x0, x1, y0, y1)
else:
rr, cc = polygon(np.array(self.draw_coordinates)[:, 1], np.array(self.draw_coordinates)[:, 0], shape=self.mask.shape)
self.mask[rr, cc] = np.maximum(self.mask[rr, cc], 255)
self.mask = self.mask.copy()
self.canvas.delete(self.current_line)
self.draw_coordinates.clear()
self.update_display()
[docs]
def finish_drawing_if_active(self, event):
if self.drawing and len(self.draw_coordinates) > 2:
self.finish_drawing(event)
####################################################################################################
# Single function butons#
####################################################################################################
[docs]
def apply_normalization(self):
self.lower_quantile.set(self.lower_entry.get())
self.upper_quantile.set(self.upper_entry.get())
self.update_display()
[docs]
def fill_objects(self):
binary_mask = self.mask > 0
filled_mask = binary_fill_holes(binary_mask)
self.mask = filled_mask.astype(np.uint8) * 255
labeled_mask, _ = label(filled_mask)
self.mask = labeled_mask
self.update_display()
[docs]
def relabel_objects(self):
mask = self.mask
labeled_mask, num_labels = label(mask > 0)
self.mask = labeled_mask
self.update_display()
[docs]
def clear_objects(self):
self.mask = np.zeros_like(self.mask)
self.update_display()
[docs]
def invert_mask(self):
self.mask = np.where(self.mask > 0, 0, 1)
self.relabel_objects()
self.update_display()
[docs]
def remove_small_objects(self):
try:
min_area = int(self.min_area_entry.get())
except ValueError:
print("Invalid minimum area value; using default of 100")
min_area = 100
labeled_mask, num_labels = label(self.mask > 0)
for i in range(1, num_labels + 1): # Skip background
if np.sum(labeled_mask == i) < min_area:
self.mask[labeled_mask == i] = 0 # Remove small objects
self.update_display()
[docs]
class AnnotateApp:
def __init__(self, root, db_path, src, image_type=None, channels=None, image_size=200, annotation_column='annotate', percentiles=(1, 99), measurement=None, threshold=None, normalize_channels=None, outline=None, outline_threshold_factor=1, outline_sigma=1, edge_thickness=1, edge_transparency=100, edge_image=False, object_size=(0,0)):
#self.update_queue.put(self.SENTINEL)
[docs]
self.SENTINEL = object() # unique sentinel for shutdown
#self.update_queue = Queue() # create the queue
if isinstance(image_size, list):
self.image_size = (int(image_size[0]), int(image_size[0]))
elif isinstance(image_size, int):
self.image_size = (image_size, image_size)
else:
raise ValueError("Invalid image size")
[docs]
self.orig_annotation_columns = annotation_column
[docs]
self.annotation_column = annotation_column
self._ensure_annotation_column()
[docs]
self.image_type = image_type
[docs]
self.channels = channels
[docs]
self.percentiles = percentiles
[docs]
self.pending_updates = {}
[docs]
self.adjusted_to_original_paths = {}
[docs]
self.update_queue = Queue()
[docs]
self.measurement = measurement
[docs]
self.threshold = threshold
[docs]
self.normalize_channels = normalize_channels
[docs]
self.outline_threshold_factor = outline_threshold_factor
[docs]
self.outline_sigma = outline_sigma
[docs]
self.edge_thickness = edge_thickness
[docs]
self.edge_transparency = edge_transparency
[docs]
self.edge_image = edge_image
[docs]
self.object_size = tuple(object_size) if object_size else (0, 0)
style_out = set_dark_style(ttk.Style())
[docs]
self.font_loader = style_out['font_loader']
[docs]
self.font_size = style_out['font_size']
[docs]
self.bg_color = style_out['bg_color']
[docs]
self.fg_color = style_out['fg_color']
[docs]
self.active_color = style_out['active_color']
[docs]
self.inactive_color = style_out['inactive_color']
# --- save-status UI & state ---
self._spinner_frames = ["â ‹","â ™","â ¸","â ´","â ¦","â ‡"] # simple TTY spinner
self._spinner_idx = 0
[docs]
self.worker_busy = False # set by the worker thread only
self._last_save_ts = None # time of last successful commit
if self.font_loader:
self.font_style = self.font_loader.get_font(size=self.font_size)
else:
self.font_style = ("Arial", 12)
self.root.configure(bg=style_out['inactive_color'])
[docs]
self.filtered_paths_annotations = []
self.prefilter_paths_annotations()
[docs]
self.db_update_thread = threading.Thread(target=self.update_database_worker)
self.db_update_thread.start()
# Set the initial window size and make it fit the screen size
self.root.geometry(f"{self.root.winfo_screenwidth()}x{self.root.winfo_screenheight()}")
self.root.update_idletasks()
# grid at top
[docs]
self.grid_frame = Frame(root, bg=self.root.cget('bg'))
self.grid_frame.grid(row=0, column=0, columnspan=2, padx=0, pady=0, sticky="nsew")
# status (left) + buttons (right) on the same bottom row
[docs]
self.status_label = Label(root, text="", font=self.font_style, bg=self.bg_color, fg=self.fg_color)
self.status_label.grid(row=2, column=0, padx=10, pady=8, sticky="w")
# begin polling the status 6–10 times/sec
self._poll_save_status() # schedules itself via .after()
self.button_frame.grid(row=2, column=1, padx=10, pady=8, sticky="e") # or "ew"
# buttons
#self.train_button = Button(self.button_frame, text="Train & Classify (beta)", command=self.train_and_classify, bg=self.bg_color, fg=self.fg_color, highlightbackground=self.fg_color, highlightcolor=self.fg_color, highlightthickness=1)
#self.umap_button = Button(self.button_frame, text="Image UMAP / HParam (Beta)", command=self.open_umap_window, bg=self.bg_color, fg=self.fg_color, highlightbackground=self.fg_color, highlightcolor=self.fg_color, highlightthickness=1)
# pack (right to left)
self.next_button.pack(side="right", padx=5)
self.previous_button.pack(side="right", padx=5)
self.exit_button.pack(side="right", padx=5)
#self.train_button.pack(side="right", padx=5)
self.settings_button.pack(side="right", padx=5)
self.clear_button.pack(side="right", padx=5)
self.count_button.pack(side="right", padx=5)
self.dl_train_button.pack(side="right", padx=5)
#self.umap_button.pack(side="right", padx=5)
# compute grid size (after buttons exist with real height)
self.button_frame.update_idletasks()
needed = self.button_frame.winfo_reqwidth()
self.root.grid_columnconfigure(1, minsize=needed + 10, weight=0)
self.root.grid_columnconfigure(0, weight=1)
self.root.update_idletasks()
self.calculate_grid_dimensions()
for i in range(self.grid_rows * self.grid_cols):
label = Label(self.grid_frame, bg=self.root.cget('bg'))
label.grid(row=i // self.grid_cols, column=i % self.grid_cols, padx=2, pady=2, sticky="nsew")
self.labels.append(label)
# column/row weights
self.root.grid_rowconfigure(0, weight=1) # grid grows
self.root.grid_rowconfigure(2, weight=0) # bottom fixed
for row in range(self.grid_rows):
self.grid_frame.grid_rowconfigure(row, weight=1)
for col in range(self.grid_cols):
self.grid_frame.grid_columnconfigure(col, weight=1)
def _int_to_color(self, k, s=0.65, v=0.95):
"""
Deterministically map any non-negative integer k -> hex color using
the golden-ratio conjugate to distribute hues around the color wheel.
s,v control saturation and value (brightness).
"""
import colorsys
# Golden ratio conjugate (~0.618...) spreads hues evenly
phi = 0.618033988749895
# Wrap k around the unit interval
h = (k * phi) % 1.0
r, g, b = colorsys.hsv_to_rgb(h, float(s), float(v))
return "#{:02x}{:02x}{:02x}".format(int(r * 255 + 0.5),
int(g * 255 + 0.5),
int(b * 255 + 0.5))
def _label_to_color(self, val):
"""
Public helper: for an integer label return a hex color.
- None/0/invalid -> None (no border).
- 1 -> blue, 2 -> red.
- 3+ -> infinite distinct colors, deterministic.
Caches results so colors stay stable across the session.
"""
# Lazy-init cache on the instance
if not hasattr(self, "_class_color_cache"):
self._class_color_cache = {}
try:
if val is None:
return None
iv = int(val)
if iv <= 0:
return None
except Exception:
return None
if iv in self._class_color_cache:
return self._class_color_cache[iv]
# Fixed starters
if iv == 1:
color = "#1f77b4" # blue
elif iv == 2:
color = "#d62728" # red
else:
# Map 3 -> k=0, 4 -> k=1, ... so early classes are well separated
k = iv - 3
color = self._int_to_color(k)
self._class_color_cache[iv] = color
return color
def _embed_figure_in(self, parent, fig):
# Clear parent
for w in parent.winfo_children():
try: w.destroy()
except Exception: pass
try:
from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg
except Exception as e:
lab = tk.Label(parent, text=f"Matplotlib not available: {e}", bg=self.bg_color, fg="red")
lab.pack()
return None
canvas = FigureCanvasTkAgg(fig, master=parent)
widget = canvas.get_tk_widget()
widget.pack(fill="both", expand=True)
canvas.draw()
return canvas
[docs]
def open_umap_window(self):
import tkinter as tk
from tkinter import ttk, messagebox
import threading
import ast
win = tk.Toplevel(self.root)
win.title("Image UMAP & Hyperparameter Search")
win.configure(bg=self.bg_color)
win.geometry("1200x800")
outer = tk.Frame(win, bg=self.bg_color)
outer.pack(fill=tk.BOTH, expand=True)
# Left: settings
left = tk.Frame(outer, bg=self.bg_color)
left.pack(side=tk.LEFT, fill=tk.Y, padx=10, pady=10)
# Right: live plot
right = tk.Frame(outer, bg=self.bg_color)
right.pack(side=tk.LEFT, fill=tk.BOTH, expand=True, padx=(0,10), pady=10)
# --- Settings widgets (a practical subset; add more if you like) ---
def _row(lbl, widget):
r = tk.Frame(left, bg=self.bg_color)
tk.Label(r, text=lbl, bg=self.bg_color, fg=self.fg_color, font=self.font_style).pack(side=tk.TOP, anchor="w")
widget.pack(in_=r, fill=tk.X, expand=True)
r.pack(fill=tk.X, pady=6)
src_entry = tk.Entry(left)
src_entry.insert(0, self.src)
tables_entry = tk.Entry(left)
tables_entry.insert(0, "cell,cytoplasm,nucleus,pathogen") # sensible default
row_limit_entry = tk.Entry(left)
row_limit_entry.insert(0, "") # blank = no limit
# UMAP params
n_neighbors_entry = tk.Entry(left); n_neighbors_entry.insert(0, "15")
min_dist_entry = tk.Entry(left); min_dist_entry.insert(0, "0.1")
metric_entry = tk.Entry(left); metric_entry.insert(0, "euclidean")
# Clustering params
clustering_cbx = ttk.Combobox(left, state="readonly", values=["dbscan","kmeans"])
clustering_cbx.set("dbscan")
eps_entry = tk.Entry(left); eps_entry.insert(0, "0.5") # DBSCAN
min_samples_entry = tk.Entry(left); min_samples_entry.insert(0, "5")
kmeans_k_entry = tk.Entry(left); kmeans_k_entry.insert(0, "8") # KMeans
color_by_entry = tk.Entry(left); color_by_entry.insert(0, "") # e.g. columnID or cond
dot_size_entry = tk.Entry(left); dot_size_entry.insert(0, "6")
fig_size_entry = tk.Entry(left); fig_size_entry.insert(0, "10") # inches
img_nr_entry = tk.Entry(left); img_nr_entry.insert(0, "200") # images in overlay (if plotting images)
plot_images_var = tk.BooleanVar(value=False)
tk.Checkbutton(left, text="plot_images (heavy)", variable=plot_images_var,
bg=self.bg_color, fg=self.fg_color, selectcolor=self.bg_color).pack(anchor="w", pady=(4,0))
# Hyperparam grids (comma-separated lists interpreted as Python literals)
# UMAP grid: list of dicts like {"n_neighbors": 10, "min_dist": 0.1}
red_grid_entry = tk.Entry(left)
red_grid_entry.insert(0, """[{"n_neighbors":10,"min_dist":0.05},{"n_neighbors":15,"min_dist":0.1},{"n_neighbors":30,"min_dist":0.3}]""")
# DBSCAN grid: list of dicts like {"eps": 0.5, "min_samples":5}
dbscan_grid_entry = tk.Entry(left)
dbscan_grid_entry.insert(0, """[{"eps":0.3,"min_samples":5},{"eps":0.5,"min_samples":5},{"eps":0.7,"min_samples":3}]""")
# KMeans grid: list of dicts like {"n_clusters": 6}
kmeans_grid_entry = tk.Entry(left)
kmeans_grid_entry.insert(0, """[{"n_clusters":6},{"n_clusters":8},{"n_clusters":10}]""")
# pack rows
_row("src", src_entry)
_row("tables (csv)", tables_entry)
_row("row_limit (blank = all)", row_limit_entry)
ttk.Separator(left, orient="horizontal").pack(fill=tk.X, pady=6)
_row("UMAP n_neighbors", n_neighbors_entry)
_row("UMAP min_dist", min_dist_entry)
_row("metric", metric_entry)
ttk.Separator(left, orient="horizontal").pack(fill=tk.X, pady=6)
_row("clustering", clustering_cbx)
_row("DBSCAN eps", eps_entry)
_row("DBSCAN min_samples", min_samples_entry)
_row("KMeans n_clusters", kmeans_k_entry)
ttk.Separator(left, orient="horizontal").pack(fill=tk.X, pady=6)
_row("color_by (optional)", color_by_entry)
_row("dot_size", dot_size_entry)
_row("figsize (inches)", fig_size_entry)
_row("image_nr (if plotting images)", img_nr_entry)
ttk.Separator(left, orient="horizontal").pack(fill=tk.X, pady=6)
_row("UMAP grid (JSON list of dicts)", red_grid_entry)
_row("DBSCAN grid (JSON list of dicts)", dbscan_grid_entry)
_row("KMeans grid (JSON list of dicts)", kmeans_grid_entry)
# Status + buttons
status = tk.Label(left, text="", bg=self.bg_color, fg=self.fg_color, font=self.font_style)
status.pack(fill=tk.X, pady=(8,4))
btn_row = tk.Frame(left, bg=self.bg_color)
btn_row.pack(fill=tk.X, pady=(0,6))
run_umap_btn = ttk.Button(btn_row, text="Run UMAP")
run_grid_btn = ttk.Button(btn_row, text="Run Hyperparam Search")
run_umap_btn.pack(side=tk.LEFT, expand=True, fill=tk.X, padx=(0,4))
run_grid_btn.pack(side=tk.LEFT, expand=True, fill=tk.X, padx=(4,0))
# ----- runners -----
def _collect_common_settings():
# parse helpers
def _csv_list(s):
s = (s or "").strip()
if not s: return []
return [p.strip() for p in s.split(",") if p.strip()]
def _int_or_none(s):
s = (s or "").strip()
return None if s == "" else int(float(s))
def _float(s, default):
try: return float(str(s).strip())
except Exception: return default
tables = _csv_list(tables_entry.get())
row_limit = _int_or_none(row_limit_entry.get())
settings = {
"src": src_entry.get().strip(),
"tables": tables if tables else ["cell","cytoplasm","nucleus","pathogen"],
"row_limit": row_limit,
"reduction_method": "umap",
"n_neighbors": _int_or_none(n_neighbors_entry.get()) or 15,
"min_dist": _float(min_dist_entry.get(), 0.1),
"metric": metric_entry.get().strip() or "euclidean",
"clustering": clustering_cbx.get().strip().upper(), # DBSCAN or KMEANS
"eps": _float(eps_entry.get(), 0.5),
"min_samples": _int_or_none(min_samples_entry.get()) or 5,
"image_nr": _int_or_none(img_nr_entry.get()) or 200,
"dot_size": _int_or_none(dot_size_entry.get()) or 6,
"figuresize": _float(fig_size_entry.get(), 10.0),
"plot_images": bool(plot_images_var.get()),
"color_by": (color_by_entry.get().strip() or None),
# defaults you already support in set_default_umap_image_settings:
"verbose": True,
"black_background": False,
"remove_image_canvas": False,
"plot_outlines": False,
"plot_points": True,
"smooth_lines": False,
"embedding_by_controls": False,
"exclude": [],
"save_figure": False,
"plot_cluster_grids": False,
"analyze_clusters": False,
"n_jobs": max(1, (os.cpu_count() or 8) - 2),
# clustering-specific extra:
"kmeans_k": _int_or_none(kmeans_k_entry.get()) or 8,
}
return settings
def _run_umap():
settings = _collect_common_settings()
def worker():
try:
status.config(text="Running UMAP…")
# Call your function; ask it to return a Figure (see tweak below)
from spacr.core import generate_image_umap as _gen
fig = _gen(settings=settings, return_fig=True)
status.config(text="Done.")
self._embed_figure_in(right, fig)
except Exception as e:
status.config(text=f"Error: {e}")
threading.Thread(target=worker, daemon=True).start()
def _run_grid():
# parse JSON-ish lists safely
def _parse_list(s):
txt = (s or "").strip()
if not txt: return []
try:
return ast.literal_eval(txt)
except Exception:
return []
settings = _collect_common_settings()
red_grid = _parse_list(red_grid_entry.get())
dbscan_grid = _parse_list(dbscan_grid_entry.get())
kmeans_grid = _parse_list(kmeans_grid_entry.get())
def worker():
try:
status.config(text="Running hyperparameter search…")
from spacr.core import reducer_hyperparameter_search as _search
fig = _search(
settings=settings,
reduction_params=red_grid or [{"n_neighbors":15,"min_dist":0.1}],
dbscan_params=dbscan_grid or [{"eps":0.5,"min_samples":5}],
kmeans_params=kmeans_grid or [{"n_clusters":settings["kmeans_k"]}],
show=False, return_fig=True
)
status.config(text="Done.")
self._embed_figure_in(right, fig)
except Exception as e:
status.config(text=f"Error: {e}")
threading.Thread(target=worker, daemon=True).start()
run_umap_btn.configure(command=_run_umap)
run_grid_btn.configure(command=_run_grid)
def _poll_save_status(self):
"""
Main-thread UI poller: reads thread-safe flags and queue length
to show a 'Saving…' spinner and counts. Never called from worker.
"""
try:
qlen = self.update_queue.qsize()
except NotImplementedError:
qlen = 0 # some platforms don't implement qsize reliably
saving = self.worker_busy or qlen > 0 or bool(self.pending_updates)
if saving:
self._spinner_idx = (self._spinner_idx + 1) % len(self._spinner_frames)
spin = self._spinner_frames[self._spinner_idx]
msg = f"{spin} Saving… queue={qlen}"
else:
if self._last_save_ts:
msg = f"✓ All changes saved"
else:
msg = "" # nothing saved yet, keep bar clean
# Update the status label in the main thread
self.status_label.config(text=msg)
# poll ~8 times per second
self.root.after(125, self._poll_save_status)
[docs]
def open_settings_window(self):
from .gui_utils import generate_annotate_fields, convert_to_number
# Create settings window
settings_window = tk.Toplevel(self.root)
settings_window.title("Modify Annotation Settings")
style_out = set_dark_style(ttk.Style())
settings_window.configure(bg=style_out['bg_color'])
settings_frame = tk.Frame(settings_window, bg=style_out['bg_color'])
settings_frame.pack(fill=tk.BOTH, expand=True)
# Generate fields with current settings pre-filled
vars_dict = generate_annotate_fields(settings_frame)
# Pre-fill the current settings into vars_dict
current_settings = {
'image_type': self.image_type or '',
'channels': ','.join(self.channels) if self.channels else '',
'img_size': f"{self.image_size[0]},{self.image_size[1]}",
'annotation_column': self.annotation_column or '',
'percentiles': ','.join(map(str, self.percentiles)),
'measurement': ','.join(self.measurement) if self.measurement else '',
'threshold': str(self.threshold) if self.threshold is not None else '',
'normalize_channels': ','.join([s for s in (self.normalize_channels or []) if isinstance(s, str) and s.strip()]),
'outline': ','.join(self.outline) if self.outline else '',
'outline_threshold_factor': str(self.outline_threshold_factor) if hasattr(self, 'outline_threshold_factor') else '1.0',
'outline_sigma': str(self.outline_sigma) if hasattr(self, 'outline_sigma') else '1.0',
'edge_thickness': str(self.edge_thickness) if hasattr(self, 'edge_thickness') else '1',
'edge_transparency': str(getattr(self, 'edge_transparency', 0.0)),
'edge_image': str(getattr(self, 'edge_image', False)),
'object_size': (f"{getattr(self, 'object_size', (0,0))[0]},{getattr(self, 'object_size', (0,0))[1]}"),
'src': self.src,
'db_path': self.db_path,
}
for key, data in vars_dict.items():
if key in current_settings:
data['entry'].delete(0, tk.END)
data['entry'].insert(0, current_settings[key])
def apply_new_settings():
settings = {key: data['entry'].get() for key, data in vars_dict.items()}
# --- channels (comma string -> list or None) ---
settings['channels'] = (
[s.strip().lower() for s in (settings.get('channels') or '').split(',') if s.strip()]
or None
)
# --- image size & percentiles ---
settings['img_size'] = list(map(int, settings['img_size'].split(',')))
settings['percentiles'] = (
list(map(convert_to_number, settings['percentiles'].split(',')))
if settings['percentiles'] else [1, 99]
)
# --- normalize_channels: empty => [] (so update_settings will process it) ---
raw_nc = settings.get('normalize_channels')
if raw_nc is None or raw_nc.strip() == '':
settings['normalize_channels'] = []
else:
nc = [s.strip().lower() for s in raw_nc.split(',') if s.strip()]
nc = [s for s in nc if s in {'r','g','b'}]
settings['normalize_channels'] = nc
raw_ol = settings.get('outline')
if raw_ol is None or raw_ol.strip() == '':
settings['outline'] = []
else:
ol = [s.strip().lower() for s in raw_ol.split(',') if s.strip()]
ol = [s for s in ol if s in {'r','g','b'}]
settings['outline'] = ol
# --- numeric fields ---
# --- object_size: "(min,max)" where 0 disables that bound ---
raw_os = (settings.get('object_size') or '').strip()
def _parse_object_size(s):
if not s:
return (0, 0)
# accept "min,max", "min , max", "min", "min," etc.
s = s.replace(';', ',')
parts = [p.strip() for p in s.split(',') if p.strip() != '']
nums = []
for p in parts[:2]:
try:
# allow floats in UI, cast to int; clamp negatives to 0
nums.append(max(0, int(float(p))))
except Exception:
nums.append(0)
while len(nums) < 2:
nums.append(0)
mn, mx = nums[0], nums[1]
# if both set and out of order, swap
if mn and mx and mn > mx:
mn, mx = mx, mn
return (mn, mx)
settings['object_size'] = _parse_object_size(raw_os)
settings['outline_threshold_factor'] = (
float(settings['outline_threshold_factor'].replace(',', '.'))
if settings['outline_threshold_factor'] else 1.0)
settings['outline_sigma'] = (
float(settings['outline_sigma'].replace(',', '.'))
if settings['outline_sigma'] else 1.0)
settings['edge_thickness'] = (
float(settings['edge_thickness'].replace(',', '.'))
if settings['edge_thickness'] else 1)
et = settings.get('edge_transparency')
if et is None or et == '':
settings['edge_transparency'] = 0.0
else:
try:
settings['edge_transparency'] = float(str(et).replace(',', '.'))
except Exception:
settings['edge_transparency'] = 0.0
settings['edge_transparency'] = max(0.0, min(100.0, settings['edge_transparency']))
# --- edge_image (string -> bool) ---
ei_raw = str(settings.get('edge_image', 'true')).strip().lower()
settings['edge_image'] = ei_raw in ('1', 'true', 't', 'yes', 'y')
# --- measurement / threshold ---
try:
settings['measurement'] = (
[s.strip() for s in settings['measurement'].split(',') if s.strip()]
if settings['measurement'] else None
)
settings['threshold'] = (
None if str(settings['threshold']).strip().lower() == 'none'
else int(settings['threshold'])
)
except Exception:
settings['measurement'] = None
settings['threshold'] = None
# --- cleanup: drop empties inside lists; only top-level '' -> None ---
for k, v in list(settings.items()):
if isinstance(v, list):
settings[k] = [x for x in v if x not in (None, '')]
elif v == '':
settings[k] = None
# db path & apply
self.db_path = os.path.join(settings.get('src'), 'measurements', 'measurements.db')
self.update_settings(**{
'image_type': settings.get('image_type'),
'channels': settings.get('channels'),
'image_size': settings.get('img_size'),
'annotation_column': settings.get('annotation_column'),
'percentiles': settings.get('percentiles'),
'measurement': settings.get('measurement'),
'threshold': settings.get('threshold'),
'normalize_channels': settings.get('normalize_channels'), # None => no normalization
'outline': settings.get('outline'), # None => no outlines
'outline_threshold_factor': settings.get('outline_threshold_factor'),
'outline_sigma': settings.get('outline_sigma'),
'edge_thickness': settings.get('edge_thickness'),
'edge_transparency': settings.get('edge_transparency'),
'edge_image': settings.get('edge_image'),
'object_size': settings.get('object_size'),
'src': settings.get('src'),
'db_path': self.db_path
})
settings_window.destroy()
apply_button = spacrButton(settings_window, text="Apply Settings", command=apply_new_settings,show_text=False, icon_name="annotate")
apply_button.pack(pady=10)
def _ensure_annotation_column(self):
import sqlite3
if not getattr(self, "annotation_column", None):
return
col = (self.annotation_column or "").replace('"', '""')
with sqlite3.connect(self.db_path, timeout=30) as conn:
cur = conn.cursor()
cur.execute('PRAGMA table_info("png_list")')
cols = {row[1] for row in cur.fetchall()}
if self.annotation_column not in cols:
# NULL allowed; values will be 1/2 per your app
cur.execute(f'ALTER TABLE "png_list" ADD COLUMN "{col}" INTEGER')
# commit occurs automatically on exiting the context if no exception
[docs]
def update_settings(self, **kwargs):
import threading
allowed_attributes = {
'image_type', 'channels', 'image_size', 'annotation_column', 'src', 'db_path',
'percentiles', 'measurement', 'threshold', 'normalize_channels',
'outline', 'outline_threshold_factor', 'outline_sigma',
'edge_thickness', 'edge_transparency', 'edge_image', 'object_size'
}
old_db = getattr(self, 'db_path', None)
old_src = getattr(self, 'src', None)
updated = False
for attr, value in kwargs.items():
if attr in allowed_attributes and value is not None:
if attr == 'normalize_channels':
if isinstance(value, (list, tuple)):
value = [str(s).strip().lower() for s in value if s is not None and str(s).strip()]
value = [s for s in value if s in {'r','g','b'}]
value = value or None
elif isinstance(value, str):
parts = [s.strip().lower() for s in value.split(',') if s.strip()]
parts = [s for s in parts if s in {'r','g','b'}]
value = parts or None
else:
value = None
elif attr == 'outline':
if isinstance(value, (list, tuple)):
value = [str(s).strip().lower() for s in value if s is not None and str(s).strip()]
elif isinstance(value, str):
value = [s.strip().lower() for s in value.split(',') if s.strip()]
else:
value = []
value = [s for s in value if s in {'r','g','b'}]
value = value or None
elif attr == 'outline_threshold_factor':
value = float(value)
elif attr == 'outline_sigma':
value = float(value)
# **CHANGED: keep fractional thickness**
elif attr == 'edge_thickness':
value = float(value)
elif attr == 'edge_transparency':
try:
value = float(value)
except Exception:
value = 0.0
value = max(0.0, min(100.0, value))
elif attr == 'edge_image':
value = bool(value)
elif attr == 'object_size':
# normalize to a 2-tuple of non-negative ints; (0,0) means no bounds
v = value
if v in (None, '', []):
v = (0, 0)
elif isinstance(v, str):
# reuse the same parsing logic as above, inline:
s = v.replace(';', ',')
parts = [p.strip() for p in s.split(',') if p.strip() != '']
a = []
for p in parts[:2]:
try:
a.append(max(0, int(float(p))))
except Exception:
a.append(0)
while len(a) < 2:
a.append(0)
mn, mx = a
elif isinstance(v, (list, tuple)):
mn = max(0, int(v[0])) if len(v) > 0 else 0
mx = max(0, int(v[1])) if len(v) > 1 else 0
else:
mn, mx = (0, 0)
if mn and mx and mn > mx:
mn, mx = mx, mn
value = (mn, mx)
setattr(self, attr, value)
updated = True
if ('annotation_column' in kwargs and kwargs['annotation_column']) or ('db_path' in kwargs and kwargs['db_path']):
self._ensure_annotation_column()
if 'image_size' in kwargs:
if isinstance(self.image_size, list):
self.image_size = (int(self.image_size[0]), int(self.image_size[0]))
elif isinstance(self.image_size, int):
self.image_size = (self.image_size, self.image_size)
elif isinstance(self.image_size, tuple) and len(self.image_size) == 2:
self.image_size = tuple(map(int, self.image_size))
else:
raise ValueError("Invalid image size")
self.calculate_grid_dimensions()
self.recreate_image_grid()
if self.src != old_src:
self.adjusted_to_original_paths.clear()
self.index = 0
if self.db_path != old_db:
if self.pending_updates:
self.update_queue.put(self.pending_updates.copy())
self.pending_updates.clear()
self.update_queue.put(self.SENTINEL)
self.update_queue.join()
try:
if getattr(self, 'db_update_thread', None):
self.db_update_thread.join()
except Exception:
pass
self.terminate = False
self.worker_busy = False
self._last_save_ts = None
self.db_update_thread = threading.Thread(target=self.update_database_worker, daemon=True)
self.db_update_thread.start()
if updated:
current_index = self.index
self.prefilter_paths_annotations()
max_index = len(self.filtered_paths_annotations) - 1
self.index = min(current_index, max(0, max(len(self.filtered_paths_annotations) - self.grid_rows * self.grid_cols, 0)))
self.load_images()
[docs]
def recreate_image_grid(self):
# Remove current labels
for label in self.labels:
label.destroy()
self.labels.clear()
# Recreate the labels grid with updated dimensions
for i in range(self.grid_rows * self.grid_cols):
label = Label(self.grid_frame, bg=self.root.cget('bg'))
label.grid(row=i // self.grid_cols, column=i % self.grid_cols, padx=2, pady=2, sticky="nsew")
self.labels.append(label)
# Reconfigure grid weights
for row in range(self.grid_rows):
self.grid_frame.grid_rowconfigure(row, weight=1)
for col in range(self.grid_cols):
self.grid_frame.grid_columnconfigure(col, weight=1)
[docs]
def update_display(self):
self.prefilter_paths_annotations()
self.load_images()
[docs]
def swich_back_annotation_column(self):
self.annotation_column = self.orig_annotation_columns
self._ensure_annotation_column()
self.prefilter_paths_annotations()
self.update_display()
[docs]
def calculate_grid_dimensions(self):
self.root.update_idletasks()
w, h = self.root.winfo_width(), self.root.winfo_height()
status_h = self.status_label.winfo_height()
buttons_h = self.button_frame.winfo_height()
bottom_h = max(status_h, buttons_h) + 8 # same row => max
self.grid_cols = max(1, w // (self.image_size[0] + 4))
self.grid_rows = max(1, (h - bottom_h) // (self.image_size[1] + 4))
[docs]
def prefilter_paths_annotations(self):
from .io import _read_and_join_tables, _read_db
from .utils import is_list_of_lists
self._ensure_annotation_column()
if self.measurement and self.threshold is not None:
df = _read_and_join_tables(self.db_path)
png_list_df = _read_db(self.db_path, tables=['png_list'])[0]
png_list_df = png_list_df.set_index('prcfo')
df = df.merge(png_list_df, left_index=True, right_index=True)
df[self.annotation_column] = None
before = len(df)
if isinstance(self.threshold, int):
if isinstance(self.measurement, list):
mes = self.measurement[0]
if isinstance(self.measurement, str):
mes = self.measurement
df = df[df[f'{mes}'] == self.threshold]
if is_list_of_lists(self.measurement):
if isinstance(self.threshold, list) or is_list_of_lists(self.threshold):
if len(self.measurement) == len(self.threshold):
for idx, var in enumerate(self.measurement):
df = df[df[var[idx]] > self.threshold[idx]]
after = len(df)
elif len(self.measurement) == len(self.threshold) * 2:
th_idx = 0
for idx, var in enumerate(self.measurement):
if idx % 2 != 0:
th_idx += 1
thd = self.threshold
if isinstance(thd, list):
thd = thd[0]
df[f'threshold_measurement_{idx}'] = df[self.measurement[idx]] / df[self.measurement[idx + 1]]
print(f"mean threshold_measurement_{idx}: {np.mean(df['threshold_measurement'])}")
print(f"median threshold measurement: {np.median(df[self.measurement])}")
df = df[df[f'threshold_measurement_{idx}'] > thd]
after = len(df)
elif isinstance(self.measurement, list):
df['threshold_measurement'] = df[self.measurement[0]] / df[self.measurement[1]]
print(f"mean threshold measurement: {np.mean(df['threshold_measurement'])}")
print(f"median threshold measurement: {np.median(df[self.measurement])}")
df = df[df['threshold_measurement'] > self.threshold]
after = len(df)
self.measurement = 'threshold_measurement'
print(f'Removed: {before-after} rows, retained {after}')
else:
print(f"mean threshold measurement: {np.mean(df[self.measurement])}")
print(f"median threshold measurement: {np.median(df[self.measurement])}")
before = len(df)
if isinstance(self.threshold, str):
if self.threshold == 'q1':
self.threshold = df[self.measurement].quantile(0.1)
if self.threshold == 'q2':
self.threshold = df[self.measurement].quantile(0.2)
if self.threshold == 'q3':
self.threshold = df[self.measurement].quantile(0.3)
if self.threshold == 'q4':
self.threshold = df[self.measurement].quantile(0.4)
if self.threshold == 'q5':
self.threshold = df[self.measurement].quantile(0.5)
if self.threshold == 'q6':
self.threshold = df[self.measurement].quantile(0.6)
if self.threshold == 'q7':
self.threshold = df[self.measurement].quantile(0.7)
if self.threshold == 'q8':
self.threshold = df[self.measurement].quantile(0.8)
if self.threshold == 'q9':
self.threshold = df[self.measurement].quantile(0.9)
print(f"threshold: {self.threshold}")
df = df[df[self.measurement] > self.threshold]
after = len(df)
print(f'Removed: {before-after} rows, retained {after}')
df = df.dropna(subset=['png_path'])
if self.image_type:
before = len(df)
if isinstance(self.image_type, list):
for tpe in self.image_type:
print(f"Looking for {tpe}")
df = df[df['png_path'].str.contains(tpe)]
print(f"Found {len(df)} entries for {tpe}")
else:
df = df[df['png_path'].str.contains(self.image_type)]
after = len(df)
print(f'image_type: Removed: {before-after} rows, retained {after}')
self.filtered_paths_annotations = df[['png_path', self.annotation_column]].values.tolist()
else:
# simple SELECT branch -> use context manager
col = (self.annotation_column or "").replace('"', '""')
with sqlite3.connect(self.db_path, timeout=30) as conn:
c = conn.cursor()
if self.image_type:
c.execute(
f'SELECT png_path, "{col}" FROM "png_list" WHERE png_path LIKE ?',
(f"%{self.image_type}%",)
)
else:
c.execute(f'SELECT png_path, "{col}" FROM "png_list"')
self.filtered_paths_annotations = c.fetchall()
[docs]
def load_images(self):
for label in self.labels:
label.config(image='')
self.images = {}
paths_annotations = self.filtered_paths_annotations[self.index:self.index + self.grid_rows * self.grid_cols]
adjusted_paths = []
for path, annotation in paths_annotations:
if not path.startswith(self.src):
parts = path.split('/data/')
if len(parts) > 1:
new_path = os.path.join(self.src, 'data', parts[1])
self.adjusted_to_original_paths[new_path] = path
adjusted_paths.append((new_path, annotation))
else:
adjusted_paths.append((path, annotation))
else:
adjusted_paths.append((path, annotation))
from concurrent.futures import ThreadPoolExecutor
with ThreadPoolExecutor() as executor:
loaded_images = list(executor.map(self.load_single_image, adjusted_paths))
for i, (img, annotation) in enumerate(loaded_images):
# NEW: infinite palette
border_color = self._label_to_color(annotation)
if border_color:
img = self.add_colored_border(img, border_width=5, border_color=border_color)
from PIL import ImageTk
photo = ImageTk.PhotoImage(img)
label = self.labels[i]
self.images[label] = photo
label.config(image=photo)
path = adjusted_paths[i][0]
label.bind('<Button-1>', self.get_on_image_click(path, label, img))
label.bind('<Button-3>', self.get_on_image_click(path, label, img))
self.root.update()
[docs]
def show_class_counts(self):
import tkinter as tk
from tkinter import ttk, messagebox
if not self.annotation_column:
messagebox.showerror("Error", "No annotation column is set.")
return
self._ensure_annotation_column()
col = (self.annotation_column or "").replace('"', '""')
with sqlite3.connect(self.db_path, timeout=30) as conn:
cur = conn.cursor()
cur.execute(
f'SELECT "{col}" AS cls, COUNT(*) '
f'FROM "png_list" '
f'WHERE "{col}" IS NOT NULL '
f'GROUP BY "{col}" '
f'ORDER BY 1'
)
rows = cur.fetchall()
win = tk.Toplevel(self.root)
win.title("Class counts (all)")
win.configure(bg=self.root.cget('bg'))
frame = tk.Frame(win, bg=self.root.cget('bg'))
frame.pack(fill=tk.BOTH, expand=True, padx=10, pady=10)
tree = ttk.Treeview(frame, columns=("cls","count","color"), show="headings", height=10)
for cid, text, width in (("cls","Class",80), ("count","Count",100), ("color","Color",120)):
tree.heading(cid, text=text)
tree.column(cid, anchor="center", width=width)
tree.pack(fill=tk.BOTH, expand=True)
# Insert rows with hex color in the last column
for cls, cnt in rows:
try:
cls_int = int(cls)
except Exception:
continue
hexcol = self._label_to_color(cls_int) or ""
tree.insert('', 'end', values=(cls_int, int(cnt), hexcol))
# Simple legend of colored squares
legend = tk.Frame(win, bg=self.root.cget('bg'))
legend.pack(fill=tk.X, padx=10, pady=8)
for cls, _ in rows[:20]: # avoid over-long legends
try:
c = int(cls)
except Exception:
continue
hx = self._label_to_color(c) or "#888888"
sw = tk.Canvas(legend, width=18, height=18, highlightthickness=0, bg=self.root.cget('bg'))
sw.create_rectangle(2, 2, 16, 16, outline=hx, fill=hx)
tk.Label(legend, text=str(c), bg=self.root.cget('bg'), fg=self.fg_color).pack(side="left", padx=(2,8))
sw.pack(side="left")
[docs]
def load_single_image(self, path_annotation_tuple):
path, annotation = path_annotation_tuple
if not os.path.exists(path):
blank = Image.new('RGB', self.image_size, color=(30, 30, 30))
print(f"Could not find image: {path}")
return blank, annotation
img = Image.open(path)
# Normalize (optionally) – returns RGB ndarray-equivalent in a PIL Image
img = self.normalize_image(img, self.percentiles, self.normalize_channels)
#img = self.normalize_image(img, self.normalize, self.percentiles, self.normalize_channels)
img = img.convert('RGB')
# Keep a copy BEFORE filtering – used for edge generation and for edge_image=True blending
full_img = img
# Apply channel filter for the visible base
img = self.filter_channels(img)
if self.outline:
img = self.outline_image(
base_img=img,
full_img=full_img,
edge_sigma=self.outline_sigma,
edge_thickness=self.edge_thickness,
fill_holes=True,
object_size=getattr(self, "object_size", (0, 0))
)
img = img.resize(self.image_size)
return img, annotation
@staticmethod
[docs]
def fill_holes(mask, min_size=0):
"""
Fill holes inside True regions of a binary mask.
Args:
mask (ndarray[bool]): Binary mask where True denotes foreground.
min_size (int): Minimum hole area to fill (in pixels).
- <= 0 : fill ALL internal holes.
- > 0 : fill only holes smaller than min_size; reopen larger ones.
Returns:
ndarray[bool]: Hole-filled mask.
"""
import numpy as np
from scipy.ndimage import binary_fill_holes, label
m = mask.astype(bool)
filled = binary_fill_holes(m)
if min_size <= 0:
return filled
# Pixels that were holes and got filled
filled_holes = filled & ~m
# Reopen (unfill) holes whose area >= min_size
lbl, n = label(filled_holes)
if n == 0:
return filled
reopen = np.zeros_like(m, dtype=bool)
for i in range(1, n + 1):
if (lbl == i).sum() >= int(min_size):
reopen |= (lbl == i)
return filled & ~reopen
@staticmethod
def _filter_objects_by_area(mask, min_size=0, max_size=0):
import numpy as np
from scipy.ndimage import label
m = mask.astype(bool)
if not m.any():
return m
lbl, n = label(m)
if n == 0:
return m
counts = np.bincount(lbl.ravel())
lo = int(min_size) if int(min_size) > 0 else 0
hi = int(max_size) if int(max_size) > 0 else np.iinfo(np.int64).max
keep = np.zeros_like(counts, dtype=bool)
for i in range(1, len(counts)):
area = counts[i]
if lo <= area <= hi:
keep[i] = True
return keep[lbl]
[docs]
def outline_image(self, base_img, full_img, edge_sigma=1, edge_thickness=1, fill_holes=True, object_size=(0, 0)):
"""
Anti-aliased outlines with sub-pixel thickness that never get dimmer as they get thinner.
Uses peak normalization so outline brightness is thickness-invariant (then scaled only by edge_transparency).
Args:
base_img (PIL.Image): already filtered (visible base)
full_img (PIL.Image): normalized RGB before filtering (edge detection / underlay)
edge_sigma (float): Gaussian smoothing before thresholding
edge_thickness (float): outline thickness in output pixels (supports < 1, e.g. 0.01)
fill_holes (bool): fill internal holes in foreground masks before boundary extraction
object_size (tuple[int,int]): (min_px, max_px) connected-component area filter.
0 disables that bound.
"""
import numpy as np
from PIL import Image
from scipy.ndimage import gaussian_filter, binary_closing
from skimage.filters import threshold_otsu
from skimage.segmentation import find_boundaries
base_arr = np.asarray(base_img).copy()
full_arr = np.asarray(full_img)
if base_arr.ndim != 3 or base_arr.shape[2] != 3:
return base_img
out_img = base_arr
channel_map = {'r': 0, 'g': 1, 'b': 2}
factor = float(getattr(self, 'outline_threshold_factor', 1.0))
# global opacity 0..1 (100 => fully bright)
transp = float(getattr(self, 'edge_transparency', 0.0))
opacity_global = max(0.0, min(1.0, transp / 100.0))
outline_channels = [ch for ch in (self.outline or []) if ch in channel_map]
show_underlay = bool(getattr(self, 'edge_image', True))
if not show_underlay and outline_channels:
for ch in outline_channels:
out_img[:, :, channel_map[ch]] = 0
if opacity_global == 0.0 or not outline_channels:
from PIL import Image as _Image
return _Image.fromarray(out_img)
# Supersampling factor (AA quality; does NOT widen geometry)
SS = 8
H, W = out_img.shape[:2]
upW, upH = W * SS, H * SS
# unpack object_size bounds
try:
min_px, max_px = object_size if object_size is not None else (0, 0)
except Exception:
min_px, max_px = (0, 0)
for ch in outline_channels:
idx = channel_map[ch]
if show_underlay:
out_img[:, :, idx] = full_arr[:, :, idx]
# Smooth & threshold (original grid)
ch_sm = gaussian_filter(full_arr[:, :, idx].astype(np.float32), sigma=float(edge_sigma))
try:
otsu = threshold_otsu(ch_sm)
except Exception:
otsu = np.percentile(ch_sm, 50.0)
thr = float(min(255.0, max(0.0, otsu * factor)))
fg_mask = (ch_sm > thr)
# Bridge tiny gaps + fill internal holes
fg_mask = binary_closing(fg_mask, structure=np.ones((3, 3), dtype=bool))
if fill_holes:
fg_mask = self.fill_holes(fg_mask, min_size=0)
# Area filtering (keep only sizes within [min_px, max_px], with 0 => no bound)
if (min_px and min_px > 0) or (max_px and max_px > 0):
fg_mask = self._filter_objects_by_area(fg_mask, min_size=min_px, max_size=max_px)
# 1-px boundary (original grid)
edge = find_boundaries(fg_mask, mode='inner').astype(np.uint8)
# Supersample WITHOUT widening: keep a crisp hi-res line
edge_img = Image.fromarray((edge * 255).astype(np.uint8), mode='L')
edge_hi = edge_img.resize((upW, upH), resample=Image.NEAREST)
edge_hi_arr = np.asarray(edge_hi, dtype=np.float32) / 255.0 # {0,1} in hi-res
# Thickness mapping (output px -> hi-res px); only dilate if >= 1 px
desired = max(0.0, float(edge_thickness))
hi_radius = desired * SS
if hi_radius >= 1.0:
from skimage.morphology import dilation, disk
r_int = int(np.floor(hi_radius))
if r_int >= 1:
thick = dilation(edge_hi_arr > 0.5, disk(r_int)).astype(np.float32)
edge_hi_arr = np.maximum(edge_hi_arr, thick)
# Downsample → anti-aliased coverage (0..1)
alpha_lo = Image.fromarray((edge_hi_arr * 255).astype(np.uint8), mode='L') \
.resize((W, H), resample=Image.LANCZOS)
alpha = np.asarray(alpha_lo, dtype=np.float32) / 255.0
# NEVER-DIM: normalize to unit peak
peak = float(alpha.max())
if peak > 0:
alpha = alpha / peak
# Apply global opacity
alpha = np.clip(alpha * opacity_global, 0.0, 1.0)
# Alpha blend onto the channel
orig = out_img[:, :, idx].astype(np.float32)
blended = alpha * 255.0 + (1.0 - alpha) * orig
out_img[:, :, idx] = np.clip(blended, 0, 255).astype(np.uint8)
return Image.fromarray(out_img)
@staticmethod
[docs]
def normalize_image(img, percentiles=(1, 99), normalize_channels=None):
"""
If normalize_channels is None or [], do nothing.
Otherwise normalize only those channels (r/g/b).
"""
img_array = np.array(img)
img_array = np.clip(img_array, 0, 255)
if not normalize_channels: # None or []
return Image.fromarray(img_array.astype('uint8'))
if img_array.ndim == 2:
p2, p98 = np.percentile(img_array, percentiles)
out = rescale_intensity(img_array, in_range=(p2, p98), out_range=(0, 255))
return Image.fromarray(np.clip(out, 0, 255).astype('uint8'))
channel_map = {'r': 0, 'g': 1, 'b': 2}
out = img_array.astype(np.float32).copy()
for ch in normalize_channels:
idx = channel_map.get(str(ch).lower())
if idx is None:
continue
p2, p98 = np.percentile(out[:, :, idx], percentiles)
out[:, :, idx] = rescale_intensity(out[:, :, idx], in_range=(p2, p98), out_range=(0, 255))
return Image.fromarray(np.clip(out, 0, 255).astype('uint8'))
[docs]
def add_colored_border(self, img, border_width, border_color):
top_border = Image.new('RGB', (img.width, border_width), color=border_color)
bottom_border = Image.new('RGB', (img.width, border_width), color=border_color)
left_border = Image.new('RGB', (border_width, img.height), color=border_color)
right_border = Image.new('RGB', (border_width, img.height), color=border_color)
bordered_img = Image.new('RGB', (img.width + 2 * border_width, img.height + 2 * border_width), color=self.fg_color)
bordered_img.paste(top_border, (border_width, 0))
bordered_img.paste(bottom_border, (border_width, img.height + border_width))
bordered_img.paste(left_border, (0, border_width))
bordered_img.paste(right_border, (img.width + border_width, border_width))
bordered_img.paste(img, (border_width, border_width))
return bordered_img
[docs]
def filter_channels(self, img):
r, g, b = img.split()
if self.channels:
# normalize and sanitize input like ['R', ' g ', None] -> {'r','g'}
chset = {str(c).strip().lower() for c in self.channels if c is not None and str(c).strip()}
if 'r' not in chset:
r = r.point(lambda _: 0)
if 'g' not in chset:
g = g.point(lambda _: 0)
if 'b' not in chset:
b = b.point(lambda _: 0)
# always return RGB; never collapse to grayscale
return Image.merge("RGB", (r, g, b))
[docs]
def get_on_image_click(self, path, label, img):
from PIL import ImageTk, ImageOps
import os
def on_image_click(event):
new_annotation = 1 if event.num == 1 else (2 if event.num == 3 else None)
original_path = self.adjusted_to_original_paths.get(path, path)
if original_path in self.pending_updates and self.pending_updates[original_path] == new_annotation:
self.pending_updates[original_path] = None
new_annotation = None
else:
self.pending_updates[original_path] = new_annotation
print(f"Image {os.path.split(path)[1]} annotated: {new_annotation}")
# Remove existing 5px border then reapply with new color (if any)
img_ = img.crop((5, 5, img.width - 5, img.height - 5))
border_fill = self._label_to_color(new_annotation)
if border_fill:
img_ = ImageOps.expand(img_, border=5, fill=border_fill)
photo = ImageTk.PhotoImage(img_)
self.images[label] = photo
label.config(image=photo)
self.root.update()
return on_image_click
@staticmethod
[docs]
def update_html(text):
display(HTML(f"""
<script>
document.getElementById('unique_id').innerHTML = '{text}';
</script>
"""))
[docs]
def clear_current_annotation(self):
import sqlite3, queue
from tkinter import messagebox
# Confirm
if not messagebox.askyesno(
"Confirm",
f'This will clear all annotations in "{self.annotation_column}".'
):
return # cancel
# Ensure column exists
self._ensure_annotation_column()
# Null the entire column (context manager => fast close/unlock)
col = (self.annotation_column or "").replace('"', '""')
with sqlite3.connect(self.db_path, timeout=30) as conn:
cur = conn.cursor()
cur.execute(f'UPDATE "png_list" SET "{col}" = NULL')
# Clear any pending updates and drain the queue
self.pending_updates.clear()
try:
while True:
self.update_queue.get_nowait()
except queue.Empty:
pass
# Refresh UI (no borders now)
self.prefilter_paths_annotations()
self.load_images()
[docs]
def update_database_worker(self):
import sqlite3, queue, time
# generous busy-timeout so short locks don't blow up under load
conn = sqlite3.connect(self.db_path, timeout=30)
cur = conn.cursor()
try:
try:
cur.execute("PRAGMA journal_mode=WAL;")
cur.execute("PRAGMA synchronous=NORMAL;")
conn.commit()
except Exception:
pass
while True:
try:
item = self.update_queue.get(timeout=0.1)
except queue.Empty:
# allow graceful exit after shutdown signal
if self.terminate:
break
continue
# --- graceful shutdown path ---
if item is self.SENTINEL:
# mark the SENTINEL as done so update_queue.join() can finish
self.update_queue.task_done()
break
# --- normal batch update ---
pending_updates = item # dict: {png_path: annotation or None}
if not pending_updates:
self.update_queue.task_done()
continue
self.worker_busy = True
col = (self.annotation_column or "").replace('"', '""')
to_null = [p for p, v in pending_updates.items() if v is None]
to_set = [(int(v), p) for p, v in pending_updates.items() if v is not None]
try:
if to_null:
cur.executemany(
f'UPDATE "png_list" SET "{col}" = NULL WHERE png_path = ?',
[(p,) for p in to_null]
)
if to_set:
cur.executemany(
f'UPDATE "png_list" SET "{col}" = ? WHERE png_path = ?',
to_set
)
conn.commit()
finally:
self.worker_busy = False
self._last_save_ts = time.time()
self.update_queue.task_done()
finally:
try:
cur.close()
except Exception:
pass
conn.close()
[docs]
def shutdown(self):
# push any pending UI updates first
if self.pending_updates:
self.update_queue.put(self.pending_updates.copy())
self.pending_updates.clear()
# signal termination and sentinel
self.terminate = True
self.update_queue.put(self.SENTINEL)
# wait for ALL tasks (including the sentinel) to be marked done
self.update_queue.join()
# now the worker has exited; join without timeout
try:
self.db_update_thread.join()
except Exception:
pass
# close UI
try:
self.root.quit()
finally:
try:
self.root.destroy()
except Exception:
pass
print("Quit application")
[docs]
def next_page(self):
if self.pending_updates:
# show saving right away until worker picks it up
self.worker_busy = True
self.update_queue.put(self.pending_updates.copy())
self.pending_updates.clear()
self.index += self.grid_rows * self.grid_cols
self.prefilter_paths_annotations()
self.load_images()
[docs]
def previous_page(self):
if self.pending_updates:
self.worker_busy = True
self.update_queue.put(self.pending_updates.copy())
self.pending_updates.clear()
self.index = max(0, self.index - self.grid_rows * self.grid_cols)
self.prefilter_paths_annotations()
self.load_images()
[docs]
def update_gui_text(self, text):
self.status_label.config(text=text)
self.root.update()
[docs]
def train_and_classify(self):
"""
1) Merge data from the relevant DB tables (including png_list).
2) Collect manual annotations from png_list.<annotation_column> => 'manual_annotation'.
- 1 => class=1, 2 => class=0 (for training).
3) If only one class is present, randomly sample unannotated images as the other class.
4) Train an XGBoost model.
5) Classify *all* rows -> fill XGboost_score (prob of class=1) & XGboost_annotation (1 or 2 if high confidence).
6) Write those columns back to sqlite.
7) Refresh the UI.
"""
import sqlite3
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix
from xgboost import XGBClassifier
# Optionally, update your GUI status label
self.update_gui_text("Merging data...")
from .io import _read_and_merge_data
# (1) Merge data
merged_df, obj_df_ls = _read_and_merge_data(
locs=[self.db_path],
tables=['cell', 'cytoplasm', 'nucleus', 'pathogen', 'png_list'],
verbose=False
)
# (2) Load manual annotations from the DB (with context manager)
self._ensure_annotation_column()
colq = (self.annotation_column or "").replace('"', '""')
with sqlite3.connect(self.db_path, timeout=30) as conn:
c = conn.cursor()
c.execute(
f'SELECT png_path, "{colq}" FROM "png_list" '
f'WHERE "{colq}" IS NOT NULL'
)
annotated_rows = c.fetchall()
annot_dict = dict(annotated_rows)
merged_df['manual_annotation'] = merged_df['png_path'].map(annot_dict)
# Subset with manual labels
annotated_df = merged_df.dropna(subset=['manual_annotation']).copy()
annotated_df['manual_annotation'] = annotated_df['manual_annotation'].replace({2: 0}).astype(int)
# (3) Handle single-class scenario
class_counts = annotated_df['manual_annotation'].value_counts()
if len(class_counts) == 1:
single_class = class_counts.index[0] # 0 or 1
needed = class_counts.iloc[0]
other_class = 1 if single_class == 0 else 0
unannotated_df_all = merged_df[merged_df['manual_annotation'].isna()].copy()
if len(unannotated_df_all) == 0:
print("No unannotated rows to sample for the other class. Cannot proceed.")
self.update_gui_text("Not enough data to train (no second class).")
return
sample_size = min(needed, len(unannotated_df_all))
artificially_labeled = unannotated_df_all.sample(n=sample_size, replace=False).copy()
artificially_labeled['manual_annotation'] = other_class
annotated_df = pd.concat([annotated_df, artificially_labeled], ignore_index=True)
print(f"Only one class was present => randomly labeled {sample_size} unannotated rows as {other_class}.")
if len(annotated_df) < 2:
print("Not enough annotated data to train (need at least 2).")
self.update_gui_text("Not enough data to train.")
return
# (4) Train XGBoost
self.update_gui_text("Training XGBoost model...")
# Identify numeric columns
ignore_cols = {'png_path', 'manual_annotation'}
feature_cols = [
col for col in annotated_df.columns
if col not in ignore_cols
and (annotated_df[col].dtype == float or annotated_df[col].dtype == int)
]
X_data = annotated_df[feature_cols].fillna(0).values
y_data = annotated_df['manual_annotation'].values
X_train, X_test, y_train, y_test = train_test_split(
X_data, y_data, test_size=0.1, random_state=42
)
model = XGBClassifier(use_label_encoder=False, eval_metric='logloss')
model.fit(X_train, y_train)
preds = model.predict(X_test)
print("=== Classification Report ===")
print(classification_report(y_test, preds))
print("=== Confusion Matrix ===")
print(confusion_matrix(y_test, preds))
# (5) Classify ALL rows
all_df = merged_df.copy()
X_all = all_df[feature_cols].fillna(0).values
probs_all = model.predict_proba(X_all)[:, 1]
all_df['XGboost_score'] = probs_all
def get_annotation_from_prob(prob):
if prob > 0.9:
return 1
elif prob < 0.1:
return 0
return None
xgb_anno_col = [get_annotation_from_prob(p) for p in probs_all]
xgb_anno_col = [2 if x == 0 else x for x in xgb_anno_col] # convert 0->2
all_df['XGboost_annotation'] = xgb_anno_col
# (6) Write results back (context manager + WAL tuning)
self.update_gui_text("Updating the database with XGBoost predictions...")
with sqlite3.connect(self.db_path, timeout=30) as conn:
c = conn.cursor()
try:
c.execute("ALTER TABLE png_list ADD COLUMN XGboost_annotation INTEGER")
except sqlite3.OperationalError:
pass
try:
c.execute("ALTER TABLE png_list ADD COLUMN XGboost_score FLOAT")
except sqlite3.OperationalError:
pass
c.execute("PRAGMA journal_mode=WAL;")
c.execute("PRAGMA synchronous=NORMAL;")
for _, row in all_df.iterrows():
score_val = float(row['XGboost_score'])
anno_val = row['XGboost_annotation']
the_path = row['png_path']
if pd.isna(the_path):
continue
if pd.isna(anno_val):
c.execute("""
UPDATE png_list
SET XGboost_annotation = NULL,
XGboost_score = ?
WHERE png_path = ?
""", (score_val, the_path))
else:
c.execute("""
UPDATE png_list
SET XGboost_annotation = ?,
XGboost_score = ?
WHERE png_path = ?
""", (int(anno_val), score_val, the_path))
# switch to the new column and (optionally) refresh the view
self.annotation_column = 'XGboost_annotation'
def _get_png_list_columns(self):
"""Return all columns from png_list; caller can decide which are 'annotation'."""
import sqlite3
cols = []
with sqlite3.connect(self.db_path, timeout=30) as conn:
cur = conn.cursor()
cur.execute('PRAGMA table_info("png_list")')
for _, name, coltype, *_ in cur.fetchall():
cols.append((name, (coltype or "").upper()))
return cols
def _parse_field_value(self, key, raw):
"""
Convert string from Entry to the right Python type using defaults as a hint.
Handles bools, ints, floats, lists (comma-separated), and passthrough strings.
"""
if raw is None:
return None
s = str(raw).strip()
if s == "":
return None
low = s.lower()
# booleans
if low in ("true", "t", "1", "yes", "y", "on"):
return True
if low in ("false", "f", "0", "no", "n", "off"):
return False
# numbers
try:
if "." in s or "e" in low:
return float(s)
return int(s)
except Exception:
pass
# lists (comma-separated) for known list keys
listy = {
"classes", "annotated_classes", "class_metadata", "train_channels",
"tables", "file_metadata"
}
if key in listy or ("," in s):
# split, trim, coerce numbers if possible
out = []
for token in s.split(","):
token = token.strip()
if token == "":
continue
try:
if "." in token or "e" in token.lower():
out.append(float(token))
else:
out.append(int(token))
except Exception:
out.append(token)
return out
return s
@staticmethod
[docs]
def convert_settings_dict_for_gui(settings):
"""
Decide widget type per setting:
- 'check' => Checkbutton (bools)
- 'combo' => readonly Combobox (predefined choices)
- 'entry' => Entry (free text / numbers / lists)
Returns: {key: (kind, options, initial)}
"""
try:
from torchvision import models as torch_models
torchvision_models = sorted({name for name, obj in torch_models.__dict__.items() if callable(obj)})
except Exception:
torchvision_models = ['resnet18', 'resnet34', 'resnet50', 'densenet121', 'mobilenet_v2']
chan_list = [
'[0,1,2,3,4,5,6,7,8]',
'[0,1,2,3,4,5,6,7]',
'[0,1,2,3,4,5,6]',
'[0,1,2,3,4,5]',
'[0,1,2,3,4]',
'[0,1,2,3]',
'[0,1,2]',
'[0,1]',
'[0]',
'[0,0]'
]
variables = {}
special_cases = {
'metadata_type': ('combo', ['cellvoyager', 'cq1', 'auto', 'custom'], 'cellvoyager'),
'channels': ('combo', chan_list, '[0,1,2,3]'),
'train_channels': ('combo', ["['r','g','b']", "['r','g']", "['r','b']", "['g','b']", "['r']", "['g']", "['b']"], "['r','g','b']"),
'channel_dims': ('combo', chan_list, '[0,1,2,3]'),
'dataset_mode': ('combo', ['annotation', 'metadata', 'measurement'], 'metadata'),
'cov_type': ('combo', ['HC0', 'HC1', 'HC2', 'HC3', None], None),
'crop_mode': ('combo', ["['cell']", "['nucleus']", "['pathogen']", "['organelle']", "['cell', 'nucleus']", "['cell', 'pathogen']", "['cell', 'organelle']", "['nucleus', 'pathogen']", "['cell', 'nucleus', 'pathogen']", "['cell', 'nucleus', 'pathogen', 'organelle']"], "['cell']"),
'timelapse_mode': ('combo', ['trackpy', 'iou', 'btrack'], 'trackpy'),
'train_mode': ('combo', ['erm', 'irm'], 'erm'),
'clustering': ('combo', ['dbscan', 'kmean'], 'dbscan'),
'reduction_method': ('combo', ['umap', 'tsne'], 'umap'),
'model_name': ('combo', ['cyto', 'cyto_2', 'cyto_3', 'nuclei'], 'cyto'),
'regression_type': ('combo', ['ols','gls','wls','rlm','glm','mixed','quantile','logit','probit','poisson','lasso','ridge'], 'ols'),
'timelapse_objects': ('combo', ["['cell']", "['nucleus']", "['pathogen']", "['cell', 'nucleus']", "['cell', 'pathogen']", "['nucleus', 'pathogen']", "['cell', 'nucleus', 'pathogen']", None], None),
'model_type': ('combo', torchvision_models, 'resnet50'),
'optimizer_type': ('combo', ['adamw', 'adam'], 'adamw'),
'schedule': ('combo', ['cosine','reduce_lr_on_plateau', 'step_lr'], 'cosine'),
'loss_type': ('combo', ['focal_loss', 'binary_cross_entropy_with_logits'], 'focal_loss'),
'normalize_by': ('combo', ['fov', 'png'], 'png'),
'agg_type': ('combo', ['mean', 'median'], 'mean'),
'grouping': ('combo', ['mean', 'median'], 'mean'),
'min_max': ('combo', ['allq', 'all'], 'allq'),
'transform': ('combo', ['log', 'sqrt', 'square', None], None),
'organelle_morphology': ('combo', ['spots', 'network', 'irregular', 'ring'], 'spots'),
'organelle_method': ('combo', ['otsu', 'adaptive', 'log', 'dog', 'ridge', 'hysteresis', 'cellpose', 'stardist', 'unet'], 'otsu'),
'organelle_model_name': ('combo', ['cyto', 'cyto2', 'cyto3', 'nuclei'], 'cyto3'),
'organelle_ridge_filter': ('combo', ['frangi', 'sato', 'meijering'], 'frangi'),
'organelle_network_threshold': ('combo', ['otsu', 'adaptive'], 'otsu'),
'organelle_ring_fill_method': ('combo', ['flood', 'convex'], 'flood'),
'organelle_stardist_model': ('combo', ['2D_versatile_fluo', '2D_paper_dsb2018'], '2D_versatile_fluo'),
'summarize_organelles_by': ('combo', ["['cell']","['nucleus']","['pathogen']","['cytoplasm']","['cell', 'nucleus']","['cell', 'pathogen']","['cell', 'cytoplasm']","['cell', 'nucleus', 'pathogen']","['cell', 'nucleus', 'pathogen', 'cytoplasm']",None], None)
}
for key, value in settings.items():
if key in special_cases:
variables[key] = special_cases[key]
elif isinstance(value, bool):
variables[key] = ('check', None, value)
elif isinstance(value, (int, float)):
variables[key] = ('entry', None, value)
elif isinstance(value, list):
variables[key] = ('entry', None, str(value))
else: # str / None / other
variables[key] = ('entry', None, "" if value is None else value)
return variables
[docs]
def build_multi_annotation(self, source_columns, target_column="multi_annot"):
"""
Consolidate multiple {1,2,NULL} columns into a single integer code column.
Unique per combination. If all inputs NULL -> store NULL.
Sets self.annotation_column to target_column and refreshes UI.
Encoding:
per col: NULL->0, 1->1, 2->2
code = 1 + sum( digit_i * (3^i) ), i = 0..N-1
(all digits 0) => store NULL instead of 1
"""
import sqlite3
if not source_columns or not isinstance(source_columns, (list, tuple)):
raise ValueError("build_multi_annotation: provide a non-empty list of source columns")
# precompute multipliers 3^i in Python (SQLite lacks POWER())
multipliers = [1]
for _ in range(1, len(source_columns)):
multipliers.append(multipliers[-1] * 3)
# safe identifiers
src_q = [f'"{c.replace(chr(34), chr(34)*2)}"' for c in source_columns]
tgt_q = f'"{target_column.replace(chr(34), chr(34)*2)}"'
# CASE to map each source to {0,1,2}
digits = [f"(CASE {c} WHEN 1 THEN 1 WHEN 2 THEN 2 ELSE 0 END)" for c in src_q]
# sum_i digit_i * 3^i
weighted_sum = " + ".join(f"{digits[i]} * {multipliers[i]}" for i in range(len(digits))) or "0"
# final value: NULL if all zero; else 1 + sum
final_expr = f"CASE WHEN ({weighted_sum}) = 0 THEN NULL ELSE (1 + {weighted_sum}) END"
with sqlite3.connect(self.db_path, timeout=30) as conn:
cur = conn.cursor()
# ensure all source columns exist (as INTEGER, NULL) so SQL won't fail
cur.execute('PRAGMA table_info("png_list")')
have = {row[1] for row in cur.fetchall()}
for col in source_columns:
if col not in have:
cq = col.replace('"','""')
cur.execute(f'ALTER TABLE "png_list" ADD COLUMN "{cq}" INTEGER')
# ensure target column exists
if target_column not in have:
tq = target_column.replace('"','""')
cur.execute(f'ALTER TABLE "png_list" ADD COLUMN "{tq}" INTEGER')
# compute in-place
cur.execute(f'UPDATE "png_list" SET {tgt_q} = {final_expr};')
conn.commit()
# make it the working annotation column and refresh view
self.annotation_column = target_column
self._ensure_annotation_column()
self.prefilter_paths_annotations()
self.load_images()
[docs]
def ensure_multi_annot_from_selection(self, source_columns, target_column="class_column", force_rebuild=True):
"""
If one column selected -> use it directly (no consolidation).
If >=2 selected -> build a consolidated column. If target_column already exists,
auto-bump to target_column_1, target_column_2, ... and use that actual name everywhere.
Returns the effective annotation column name.
"""
import sqlite3
if not source_columns or not isinstance(source_columns, (list, tuple)):
raise ValueError("ensure_multi_annot_from_selection: provide a non-empty list of source columns")
# Single column => just use it directly
if len(source_columns) == 1:
self.annotation_column = source_columns[0]
self._ensure_annotation_column()
self.prefilter_paths_annotations()
self.load_images()
return self.annotation_column
# Multi-column consolidation: pick a free target name (auto-bump)
with sqlite3.connect(self.db_path, timeout=30) as conn:
cur = conn.cursor()
cur.execute('PRAGMA table_info("png_list")')
existing = {row[1] for row in cur.fetchall()}
base = (str(target_column).strip() or "class_column")
effective = base
suffix = 1
while effective in existing:
effective = f"{base}_{suffix}"
suffix += 1
# Build / refresh the consolidated column under 'effective'
# (build_multi_annotation will set self.annotation_column and refresh UI)
if force_rebuild or self.annotation_column != effective:
self.build_multi_annotation(source_columns, target_column=effective)
else:
self.build_multi_annotation(source_columns, target_column=effective)
# Ensure local state reflects the chosen column name
self.annotation_column = effective
self._ensure_annotation_column()
self.prefilter_paths_annotations()
self.load_images()
return effective
[docs]
def open_deep_spacr_window(self):
import tkinter as tk
from tkinter import ttk, messagebox
import sqlite3, threading, ast, json, os
from spacr.settings import deep_spacr_defaults
# ---- defaults ---------------------------------------------------------
defaults = deep_spacr_defaults({})
defaults['src'] = self.src or defaults.get('src')
defaults['dataset'] = defaults.get('dataset', defaults['src'])
defaults['annotation_column'] = self.annotation_column or defaults.get('annotation_column')
# keep your app-wide style usage
style_out = set_dark_style(ttk.Style())
bg = self.bg_color
fg = self.fg_color
font = self.font_style
# ---- window -----------------------------------------------------------
win = tk.Toplevel(self.root)
win.title("Deep SPACR — Train")
win.configure(bg=bg)
win.geometry("1120x760")
outer = tk.Frame(win, bg=bg)
outer.pack(fill=tk.BOTH, expand=True, padx=10, pady=10)
# ---- header master toggles (govern tab enablement) --------------------
header = tk.Frame(outer, bg=bg)
header.pack(fill=tk.X, pady=(0,8))
gen_var = tk.BooleanVar(value=bool(defaults.get('generate_training_dataset', True)))
#train_var = tk.BooleanVar(value=bool(defaults.get('train_DL_model', True)))
train_var = tk.BooleanVar(value=bool(defaults.get('train', False) or defaults.get('test', False)))
apply_var = tk.BooleanVar(value=bool(defaults.get('apply_model_to_dataset', True)))
def _chk(label, var):
return tk.Checkbutton(
header, text=label, variable=var,
bg=bg, fg=fg, selectcolor=bg, font=font, activebackground=bg, activeforeground=fg
)
_chk("Generate training dataset", gen_var).pack(side=tk.LEFT, padx=(0,12))
_chk("Train", train_var).pack(side=tk.LEFT, padx=(0,12))
_chk("Apply model to dataset", apply_var).pack(side=tk.LEFT, padx=(0,12))
# ---- notebook ---------------------------------------------------------
nb = ttk.Notebook(outer)
nb.pack(fill=tk.BOTH, expand=True)
# ---- helpers ----------------------------------------------------------
def _label(parent, text):
return tk.Label(parent, text=text, bg=bg, fg=fg, anchor='w', font=font)
def _row(parent, r, label_text, widget):
_label(parent, label_text).grid(row=r, column=0, sticky="w", padx=6, pady=4)
widget.grid(row=r, column=1, sticky="ew", padx=6, pady=4)
parent.grid_columnconfigure(1, weight=1)
def _parse_list_literal(s, fallback=None):
if s is None or str(s).strip() == "":
return fallback
try:
return ast.literal_eval(str(s))
except Exception:
return fallback
def _parse_csv_list(s, fallback=None):
if s is None or str(s).strip() == "":
return fallback
parts = [p.strip() for p in str(s).split(",") if p.strip() != ""]
return parts if parts else fallback
def _set_disabled_state(frame, disabled=True):
# toggle "viability": enable/disable inputs inside the frame
state = tk.DISABLED if disabled else tk.NORMAL
for child in frame.winfo_children():
try:
child.configure(state=state)
except Exception:
pass
# ======================================================================
# TAB 1: Generate training dataset
# ======================================================================
tab_gen = tk.Frame(nb, bg=bg)
nb.add(tab_gen, text="Generate training dataset")
gen_split = tk.PanedWindow(tab_gen, orient=tk.HORIZONTAL, sashwidth=6, bg=bg)
gen_split.pack(fill=tk.BOTH, expand=True)
gen_form = tk.Frame(gen_split, bg=bg)
gen_right = tk.Frame(gen_split, bg=bg)
gen_split.add(gen_form)
gen_split.add(gen_right)
# --- Left column (general) --------------------------------------------
r = 0
dataset_mode_cbx = ttk.Combobox(gen_form, values=['annotation','metadata','measurement'], state='readonly')
dataset_mode_cbx.set(defaults.get('dataset_mode', 'annotation'))
_row(gen_form, r, "dataset_mode", dataset_mode_cbx); r += 1
size_sp = ttk.Spinbox(gen_form, from_=16, to=4096, increment=16)
size_sp.set(int(defaults.get('size', 224)))
_row(gen_form, r, "size (cropped PNG side)", size_sp); r += 1
img_size_sp = ttk.Spinbox(gen_form, from_=16, to=4096, increment=16)
img_size_sp.set(int(defaults.get('image_size', 224)))
_row(gen_form, r, "image_size (model input)", img_size_sp); r += 1
test_split_sp = ttk.Spinbox(gen_form, from_=0.0, to=0.9, increment=0.01)
test_split_sp.set(float(defaults.get('test_split', 0.1)))
_row(gen_form, r, "test_split", test_split_sp); r += 1
sample_sp = ttk.Spinbox(gen_form, from_=0, to=10**9, increment=1)
sample_val = defaults.get('sample', None)
sample_sp.delete(0, tk.END)
sample_sp.insert(0, "" if sample_val in (None, "") else str(sample_val))
_row(gen_form, r, "sample (rows, optional)", sample_sp); r += 1
# FILE TYPE: free text (any string)
file_type_entry = tk.Entry(gen_form)
file_type_entry.insert(0, str(defaults.get('file_type', defaults.get('png_type','cell_png'))))
_row(gen_form, r, "file_type / png_type", file_type_entry); r += 1
tables_entry = tk.Entry(gen_form)
tables_entry.insert(0, "" if defaults.get('tables') in (None, []) else ",".join(defaults.get('tables')))
_row(gen_form, r, "tables (csv)", tables_entry); r += 1
file_metadata_entry = tk.Entry(gen_form)
if defaults.get('file_metadata') not in (None, []):
file_metadata_entry.insert(0, ",".join(defaults['file_metadata']) if isinstance(defaults['file_metadata'], list) else str(defaults['file_metadata']))
_row(gen_form, r, "file_metadata (csv)", file_metadata_entry); r += 1
metadata_type_by_cbx = ttk.Combobox(gen_form, state='readonly', values=['columnID','something_else'])
metadata_type_by_cbx.set(defaults.get('metadata_type_by','columnID'))
_row(gen_form, r, "metadata_type_by", metadata_type_by_cbx); r += 1
class_metadata_entry = tk.Entry(gen_form)
class_metadata_entry.insert(0, str(defaults.get('class_metadata', [['c1'],['c2']])))
_row(gen_form, r, "class_metadata (list-of-lists)", class_metadata_entry); r += 1
classes_entry = tk.Entry(gen_form)
classes_entry.insert(0, str(defaults.get('classes', ['nc','pc'])))
_row(gen_form, r, "classes (list)", classes_entry); r += 1
annotated_classes_entry = tk.Entry(gen_form)
annotated_classes_entry.insert(0, str(defaults.get('annotated_classes', [1,2])))
_row(gen_form, r, "annotated_classes (list)", annotated_classes_entry); r += 1
ch_interest_sp = ttk.Spinbox(gen_form, from_=1, to=5, increment=1)
ch_interest_sp.set(int(defaults.get('channel_of_interest', 3)))
_row(gen_form, r, "channel_of_interest", ch_interest_sp); r += 1
custom_measurement_entry = tk.Entry(gen_form)
if defaults.get('custom_measurement'):
custom_measurement_entry.insert(0, str(defaults['custom_measurement']))
_row(gen_form, r, "custom_measurement (optional)", custom_measurement_entry); r += 1
balance_var = tk.BooleanVar(value=bool(defaults.get('balance_to_smallest', True)))
balance_chk = tk.Checkbutton(gen_form, text="Balance classes to smallest",
variable=balance_var, bg=bg, fg=fg, selectcolor=bg, font=font,
activebackground=bg, activeforeground=fg)
_row(gen_form, r, "", balance_chk); r += 1
# --- Right column: three MODE-SPECIFIC panels -------------------------
# 1) Annotation panel
ann_frame = tk.LabelFrame(gen_right, text="Annotation columns", bg=bg, fg=fg, font=font, labelanchor='n')
ann_inner = tk.Frame(ann_frame, bg=bg)
_label(ann_inner, "Use DB Annotation Columns").grid(row=0, column=0, sticky="w", padx=6, pady=(8,2))
use_db_var = tk.BooleanVar(value=True)
tk.Checkbutton(ann_inner, text="Use selected DB columns as classes",
variable=use_db_var, bg=bg, fg=fg, selectcolor=bg, font=font,
activebackground=bg, activeforeground=fg).grid(row=1, column=0, sticky="w", padx=6, pady=(0,6))
lb = tk.Listbox(ann_inner, selectmode=tk.EXTENDED, height=10,
bg=self.inactive_color, fg=fg, highlightbackground=fg, selectbackground=self.active_color)
lb.grid(row=2, column=0, sticky="nsew", padx=6, pady=(0,8))
ann_inner.grid_columnconfigure(0, weight=1)
ann_inner.grid_rowconfigure(2, weight=1)
try:
with sqlite3.connect(self.db_path, timeout=10) as conn:
cur = conn.cursor()
cur.execute('PRAGMA table_info("png_list")')
for _, name, coltype, *_ in cur.fetchall():
nm = str(name)
if nm.lower() in ('png_path', 'prcfo'):
continue
if (coltype or '').upper().startswith('INT') or nm not in ('png_path',):
lb.insert(tk.END, nm)
except Exception:
pass
ann_inner.pack(fill=tk.BOTH, expand=True)
# 2) Metadata panel
meta_grp = tk.LabelFrame(gen_right, text="Metadata rules (JSON)", bg=bg, fg=fg, font=font, labelanchor='n')
meta_inner = tk.Frame(meta_grp, bg=bg)
meta_rules_entry = tk.Entry(meta_inner)
meta_rules_entry.pack(fill=tk.X, padx=6, pady=(6,4))
ex = tk.Frame(meta_inner, bg=bg)
tk.Label(
ex,
text=("Example:\n"
"[\n"
" {\"name\":\"test_1\", \"where\":[{\"column\":\"test\",\"op\":\"==\",\"value\":1}]},\n"
" {\"name\":\"test_2\", \"where\":[{\"column\":\"test\",\"op\":\"==\",\"value\":2}]},\n"
" {\"name\":\"parasite_1\", \"where\":[{\"column\":\"parasite\",\"op\":\"==\",\"value\":1}]}\n"
"]"),
justify='left', anchor='w', bg=bg, fg=fg
).pack(side=tk.LEFT, fill=tk.X, expand=True)
def _insert_meta_example():
meta_rules_entry.delete(0, tk.END)
meta_rules_entry.insert(
0,
'[{"name":"test_1","where":[{"column":"test","op":"==","value":1}]},'
' {"name":"test_2","where":[{"column":"test","op":"==","value":2}]},'
' {"name":"parasite_1","where":[{"column":"parasite","op":"==","value":1}]}]'
)
ttk.Button(ex, text="Insert example", command=_insert_meta_example).pack(side=tk.RIGHT, padx=6)
ex.pack(fill=tk.X, padx=6, pady=(0,6))
meta_inner.pack(fill=tk.BOTH, expand=True)
# 3) Measurement panel
meas_grp = tk.LabelFrame(gen_right, text="Measurement selection", bg=bg, fg=fg, font=font, labelanchor='n')
meas_inner = tk.Frame(meas_grp, bg=bg)
_label(meas_inner, "measurement (csv: columns)").grid(row=0, column=0, sticky="w", padx=6, pady=(8,2))
meas_cols_entry = tk.Entry(meas_inner)
meas_cols_entry.insert(0, "" if defaults.get('measurement') in (None, []) else (
",".join(defaults['measurement']) if isinstance(defaults['measurement'], list) else str(defaults['measurement'])
))
meas_cols_entry.grid(row=0, column=1, sticky="ew", padx=6, pady=(8,2))
_label(meas_inner, "threshold (float or q1..q9)").grid(row=1, column=0, sticky="w", padx=6, pady=(4,2))
threshold_entry = tk.Entry(meas_inner)
threshold_entry.insert(0, str(defaults.get('threshold', 'q8')))
threshold_entry.grid(row=1, column=1, sticky="ew", padx=6, pady=(4,2))
tk.Label(meas_inner, text="Examples: 0.42 or q7", bg=bg, fg=fg, font=font)\
.grid(row=2, column=1, sticky="w", padx=6, pady=(0,6))
meas_inner.grid_columnconfigure(1, weight=1)
meas_inner.pack(fill=tk.BOTH, expand=True)
# start with correct panel visible & viable
def _toggle_gen_right(*_):
mode = dataset_mode_cbx.get().strip().lower()
# hide all
for w in (ann_frame, meta_grp, meas_grp):
w.pack_forget()
_set_disabled_state(w, disabled=True)
# show + enable chosen
if mode == 'annotation':
ann_frame.pack(fill=tk.BOTH, expand=True, padx=6, pady=(0,8))
_set_disabled_state(ann_frame, disabled=False)
elif mode == 'metadata':
meta_grp.pack(fill=tk.BOTH, expand=True, padx=6, pady=(0,8))
_set_disabled_state(meta_grp, disabled=False)
elif mode == 'measurement':
meas_grp.pack(fill=tk.BOTH, expand=True, padx=6, pady=(0,8))
_set_disabled_state(meas_grp, disabled=False)
dataset_mode_cbx.bind("<<ComboboxSelected>>", _toggle_gen_right)
_toggle_gen_right()
# ======================================================================
# TAB 2: Train
# ======================================================================
tab_train = tk.Frame(nb, bg=bg)
nb.add(tab_train, text="Train")
tr_basic = tk.LabelFrame(tab_train, text="Basic", bg=bg, fg=fg)
tr_basic.pack(fill=tk.X, padx=8, pady=(8,6))
rr = 0
try:
import torchvision
model_names = sorted({n for n, o in getattr(torchvision.models, '__dict__', {}).items() if callable(o)})
except Exception:
model_names = ['resnet18','resnet34','resnet50','densenet121','mobilenet_v2']
model_cbx = ttk.Combobox(tr_basic, state='readonly', values=model_names)
model_cbx.set(defaults.get('model_type', 'resnet50'))
_row(tr_basic, rr, "model_type", model_cbx); rr += 1
epochs_sp = ttk.Spinbox(tr_basic, from_=1, to=2000, increment=1)
epochs_sp.set(int(defaults.get('epochs', 100)))
_row(tr_basic, rr, "epochs", epochs_sp); rr += 1
bs_sp = ttk.Spinbox(tr_basic, from_=1, to=4096, increment=1)
bs_sp.set(int(defaults.get('batch_size', 64)))
_row(tr_basic, rr, "batch_size", bs_sp); rr += 1
lr_sp = ttk.Spinbox(tr_basic, from_=1e-6, to=1e-1, increment=1e-6)
lr_sp.set(float(defaults.get('learning_rate', 1e-3)))
_row(tr_basic, rr, "learning_rate", lr_sp); rr += 1
val_split_sp = ttk.Spinbox(tr_basic, from_=0.0, to=0.9, increment=0.01)
val_split_sp.set(float(defaults.get('val_split', 0.1)))
_row(tr_basic, rr, "val_split", val_split_sp); rr += 1
loss_cbx = ttk.Combobox(tr_basic, state='readonly',
values=["auto","ce","ce_smooth","ce_weighted","focal_ce","bce","focal_bce","logit_adjust_ce", "asl"])
loss_cbx.set(defaults.get('loss_type', 'auto'))
_row(tr_basic, rr, "loss_type", loss_cbx); rr += 1
train_channels_cbx = ttk.Combobox(tr_basic, state='readonly',
values=["['r','g','b']", "['r','g']", "['r','b']", "['g','b']", "['r']", "['g']", "['b']"])
tdef = defaults.get('train_channels', ['r','g','b'])
train_channels_cbx.set(str(tdef if isinstance(tdef, list) else "['r','g','b']"))
_row(tr_basic, rr, "train_channels", train_channels_cbx); rr += 1
do_train_var = tk.BooleanVar(value=bool(defaults.get('train', True)))
do_test_var = tk.BooleanVar(value=bool(defaults.get('test', False)))
_row(tr_basic, rr, "", tk.Checkbutton(tr_basic, text="train (legacy flag)",
variable=do_train_var, bg=bg, fg=fg, selectcolor=bg, font=font)); rr += 1
_row(tr_basic, rr, "", tk.Checkbutton(tr_basic, text="test after training (legacy flag)",
variable=do_test_var, bg=bg, fg=fg, selectcolor=bg, font=font)); rr += 1
adv = tk.LabelFrame(tab_train, text="Advanced", bg=bg, fg=fg)
adv.pack(fill=tk.X, padx=8, pady=(0,8))
ra = 0
opt_cbx = ttk.Combobox(adv, state='readonly', values=['adamw','adagrad','adam'])
opt_cbx.set(defaults.get('optimizer_type', 'adamw'))
_row(adv, ra, "optimizer_type", opt_cbx); ra += 1
sched_cbx = ttk.Combobox(adv, state='readonly', values=['cosine','reduce_lr_on_plateau','step_lr'])
sched_cbx.set(defaults.get('schedule', 'cosine'))
_row(adv, ra, "schedule", sched_cbx); ra += 1
wd_sp = ttk.Spinbox(adv, from_=0.0, to=1.0, increment=1e-6)
wd_sp.set(float(defaults.get('weight_decay', 1e-5)))
_row(adv, ra, "weight_decay", wd_sp); ra += 1
dr_sp = ttk.Spinbox(adv, from_=0.0, to=0.9, increment=0.01)
dr_sp.set(float(defaults.get('dropout_rate', 0.1)))
_row(adv, ra, "dropout_rate", dr_sp); ra += 1
init_w_var = tk.BooleanVar(value=bool(defaults.get('init_weights', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="init_weights",
variable=init_w_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
use_ckpt_var = tk.BooleanVar(value=bool(defaults.get('use_checkpoint', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="use_checkpoint (activation checkpointing)",
variable=use_ckpt_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
amsgrad_var = tk.BooleanVar(value=bool(defaults.get('amsgrad', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="AMSGrad",
variable=amsgrad_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
intermed_var = tk.BooleanVar(value=bool(defaults.get('intermedeate_save', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="intermedeate_save",
variable=intermed_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
jobs_sp = ttk.Spinbox(adv, from_=0, to=max(1, os.cpu_count() or 64), increment=1)
jobs_sp.set(int(defaults.get('n_jobs', max(1, (os.cpu_count() or 8)-4))))
_row(adv, ra, "n_jobs (DataLoader workers)", jobs_sp); ra += 1
pin_var = tk.BooleanVar(value=bool(defaults.get('pin_memory', False)))
_row(adv, ra, "", tk.Checkbutton(adv, text="pin_memory",
variable=pin_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
ga_sp = ttk.Spinbox(adv, from_=1, to=64, increment=1)
ga_sp.set(int(defaults.get('gradient_accumulation_steps', 4)))
_row(adv, ra, "gradient_accumulation_steps", ga_sp); ra += 1
grad_acc_var = tk.BooleanVar(value=bool(defaults.get('gradient_accumulation', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="gradient_accumulation",
variable=grad_acc_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
augment_var = tk.BooleanVar(value=bool(defaults.get('augment', False)))
_row(adv, ra, "", tk.Checkbutton(adv, text="augment",
variable=augment_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
normalize_var = tk.BooleanVar(value=bool(defaults.get('normalize', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="normalize",
variable=normalize_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
verbose_var = tk.BooleanVar(value=bool(defaults.get('verbose', True)))
_row(adv, ra, "", tk.Checkbutton(adv, text="verbose",
variable=verbose_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
custom_model_var = tk.BooleanVar(value=bool(defaults.get('custom_model', False)))
_row(adv, ra, "", tk.Checkbutton(adv, text="custom_model",
variable=custom_model_var, bg=bg, fg=fg, selectcolor=bg, font=font)); ra += 1
custom_model_entry = tk.Entry(adv)
custom_model_entry.insert(0, str(defaults.get('custom_model_path','path')))
_row(adv, ra, "custom_model_path", custom_model_entry); ra += 1
# ======================================================================
# TAB 3: Apply model to dataset
# ======================================================================
tab_apply = tk.Frame(nb, bg=bg)
nb.add(tab_apply, text="Apply model")
apply_frame = tk.LabelFrame(tab_apply, text="Inference", bg=bg, fg=fg)
apply_frame.pack(fill=tk.X, padx=8, pady=8)
rr2 = 0
score_sp = ttk.Spinbox(apply_frame, from_=0.0, to=1.0, increment=0.01)
score_sp.set(float(defaults.get('score_threshold', 0.5)))
_row(apply_frame, rr2, "score_threshold", score_sp); rr2 += 1
dataset_entry = tk.Entry(apply_frame)
dataset_entry.insert(0, str(defaults.get('dataset', defaults['src'])))
_row(apply_frame, rr2, "dataset (apply on this path)", dataset_entry); rr2 += 1
model_path_entry = tk.Entry(apply_frame)
model_path_entry.insert(0, str(defaults.get('model_path','path')))
_row(apply_frame, rr2, "model_path (optional override)", model_path_entry); rr2 += 1
# ---- enable/disable tabs based on header toggles ----------------------
def _apply_tab_state(*_):
nb.tab(0, state='normal' if gen_var.get() else 'disabled')
nb.tab(1, state='normal' if train_var.get() else 'disabled')
nb.tab(2, state='normal' if apply_var.get() else 'disabled')
for var in (gen_var, train_var, apply_var):
var.trace_add("write", _apply_tab_state)
_apply_tab_state()
# ---- bottom buttons ---------------------------------------------------
btns = tk.Frame(win, bg=bg)
btns.pack(fill=tk.X, padx=10, pady=(0,10))
run_btn = ttk.Button(btns, text="Run")
cancel_btn = ttk.Button(btns, text="Cancel", command=win.destroy)
run_btn.pack(side=tk.RIGHT, padx=5)
cancel_btn.pack(side=tk.RIGHT, padx=5)
# ---- run handler ------------------------------------------------------
def on_run():
settings = dict(defaults) # copy
# Master toggles
settings['generate_training_dataset'] = bool(gen_var.get())
#settings['train_DL_model'] = bool(train_var.get())
settings['apply_model_to_dataset'] = bool(apply_var.get())
# GENERATE / DATASET (shared)
mode = dataset_mode_cbx.get().strip()
settings['dataset_mode'] = mode
settings['size'] = int(float(size_sp.get()))
settings['image_size'] = int(float(img_size_sp.get()))
settings['test_split'] = float(test_split_sp.get())
settings['sample'] = None if str(sample_sp.get()).strip() == "" else int(float(sample_sp.get()))
ft = file_type_entry.get().strip()
settings['file_type'] = ft
settings['png_type'] = ft
settings['tables'] = _parse_csv_list(tables_entry.get(), None)
settings['file_metadata'] = _parse_csv_list(file_metadata_entry.get(), None)
settings['metadata_type_by'] = metadata_type_by_cbx.get().strip()
settings['class_metadata'] = _parse_list_literal(class_metadata_entry.get(), defaults.get('class_metadata'))
settings['classes'] = _parse_list_literal(classes_entry.get(), defaults.get('classes'))
settings['annotated_classes'] = _parse_list_literal(annotated_classes_entry.get(), defaults.get('annotated_classes'))
settings['channel_of_interest'] = int(float(ch_interest_sp.get()))
cm = custom_measurement_entry.get().strip()
settings['custom_measurement'] = (cm if cm != "" else None)
settings['balance_to_smallest'] = bool(balance_var.get())
# MODE-SPECIFIC
if mode == 'annotation':
settings['use_db_columns'] = bool(use_db_var.get())
if settings['use_db_columns']:
sel_cols = [lb.get(i) for i in lb.curselection()]
if not sel_cols:
messagebox.showwarning("No DB columns selected", "Select at least one annotation column or uncheck the DB option.")
return
# Build/choose effective consolidated column name.
# Base name is "class_column"; if it exists, you'll get class_column_1, _2, ...
base_name = "class_column" if len(sel_cols) > 1 else sel_cols[0]
effective_col = self.ensure_multi_annot_from_selection(
sel_cols, target_column=base_name, force_rebuild=True
)
settings['annotation_column'] = effective_col
else:
settings['annotation_column'] = self.annotation_column
# Remove non-annotation keys
settings.pop('metadata_rules', None)
settings.pop('measurement', None)
settings.pop('threshold', None)
elif mode == 'metadata':
raw = meta_rules_entry.get().strip()
rules = None
if raw:
try:
rules = json.loads(raw)
except Exception:
rules = _parse_list_literal(raw, None)
if not rules:
messagebox.showwarning("Metadata rules", "Provide valid JSON rules or click 'Insert example'.")
return
settings['metadata_rules'] = rules
settings.pop('measurement', None)
settings.pop('threshold', None)
settings.pop('annotation_column', None)
settings.pop('db_annotation_columns', None)
settings.pop('use_db_columns', None)
elif mode == 'measurement':
meas_cols = _parse_csv_list(meas_cols_entry.get(), None)
if not meas_cols:
messagebox.showwarning("Measurement", "Provide at least one measurement column (csv).")
return
settings['measurement'] = meas_cols if len(meas_cols) > 1 else meas_cols[0]
th_raw = threshold_entry.get().strip()
if th_raw == "":
messagebox.showwarning("Measurement", "Provide a threshold (number) or a quantile code q1..q9.")
return
try:
settings['threshold'] = float(th_raw)
except Exception:
settings['threshold'] = th_raw # e.g. "q8"
settings.pop('metadata_rules', None)
settings.pop('annotation_column', None)
settings.pop('db_annotation_columns', None)
settings.pop('use_db_columns', None)
# TRAIN
settings['model_type'] = model_cbx.get().strip()
settings['epochs'] = int(float(epochs_sp.get()))
settings['batch_size'] = int(float(bs_sp.get()))
settings['learning_rate'] = float(lr_sp.get())
settings['val_split'] = float(val_split_sp.get())
settings['loss_type'] = loss_cbx.get().strip()
settings['train_channels'] = _parse_list_literal(train_channels_cbx.get(), ['r','g','b'])
settings['train'] = bool(do_train_var.get()) # legacy flag
settings['test'] = bool(do_test_var.get()) # legacy flag
settings['optimizer_type'] = opt_cbx.get().strip()
settings['schedule'] = sched_cbx.get().strip()
settings['weight_decay'] = float(wd_sp.get())
settings['dropout_rate'] = float(dr_sp.get())
settings['init_weights'] = bool(init_w_var.get())
settings['use_checkpoint'] = bool(use_ckpt_var.get())
settings['amsgrad'] = bool(amsgrad_var.get())
settings['intermedeate_save'] = bool(intermed_var.get())
settings['n_jobs'] = int(float(jobs_sp.get()))
settings['pin_memory'] = bool(pin_var.get())
settings['gradient_accumulation_steps'] = int(float(ga_sp.get()))
settings['gradient_accumulation'] = bool(grad_acc_var.get())
settings['augment'] = bool(augment_var.get())
settings['normalize'] = bool(normalize_var.get())
settings['verbose'] = bool(verbose_var.get())
settings['custom_model'] = bool(custom_model_var.get())
settings['custom_model_path'] = custom_model_entry.get().strip() or settings.get('custom_model_path')
# APPLY
settings['score_threshold'] = float(score_sp.get())
settings['dataset'] = dataset_entry.get().strip() or self.src
mp = model_path_entry.get().strip()
if mp:
settings['model_path'] = mp
# Essentials
settings['src'] = self.src
win.destroy()
def _worker():
try:
self.update_gui_text("Deep SPACR: preparing…")
from spacr.deep_spacr import deep_spacr
deep_spacr(settings)
self.update_gui_text("Deep SPACR: done.")
except Exception as e:
import traceback
traceback.print_exc()
self.update_gui_text(f"Deep SPACR error: {e}")
threading.Thread(target=_worker, daemon=True).start()
run_btn.configure(command=on_run)
[docs]
def generate_dna_matrix(output_path='dna_matrix.gif', canvas_width=1500, canvas_height=1000, duration=30, fps=20, base_size=20, transition_frames=30, font_type='arial.ttf', enhance=[1.1, 1.5, 1.2, 1.5], lowercase_prob=0.3):
"""
Generate a DNA matrix animation and save it as GIF, MP4, or AVI using OpenCV for videos.
"""
def save_output(frames, output_path, fps, output_format):
"""Save the animation based on output format."""
if output_format in ['.mp4', '.avi']:
images = [np.array(img.convert('RGB')) for img in frames]
fourcc = cv2.VideoWriter_fourcc(*('mp4v' if output_format == '.mp4' else 'XVID'))
out = cv2.VideoWriter(output_path, fourcc, fps, (canvas_width, canvas_height))
for img in images:
out.write(cv2.cvtColor(img, cv2.COLOR_RGB2BGR))
out.release()
elif output_format == '.gif':
frames[0].save(output_path, save_all=True, append_images=frames[1:], duration=int(1000/fps), loop=0)
def draw_base(draw, col_idx, base_position, base, font, alpha=255, fill_color=None):
"""Draws a DNA base at the specified position."""
draw.text((col_idx * base_size, base_position * base_size), base, fill=(*fill_color, alpha), font=font)
# Setup variables
num_frames = duration * fps
num_columns = canvas_width // base_size
bases = ['A', 'T', 'C', 'G']
active_color = (155, 55, 155)
color = (255, 255, 255)
base_colors = {'A': color, 'T': color, 'C': color, 'G': color}
_, output_format = os.path.splitext(output_path)
# Initialize font
try:
font = ImageFont.truetype(font_type, base_size)
except IOError:
font = ImageFont.load_default()
# DNA string and positions
string_lengths = [random.randint(10, 100) for _ in range(num_columns)]
visible_bases = [0] * num_columns
base_positions = [random.randint(-canvas_height // base_size, 0) for _ in range(num_columns)]
column_strings = [[''] * 100 for _ in range(num_columns)]
random_white_sequences = [None] * num_columns
frames = []
end_frame_start = int(num_frames * 0.8)
for frame_idx in range(num_frames):
img = Image.new('RGBA', (canvas_width, canvas_height), color=(0, 0, 0, 255))
draw = ImageDraw.Draw(img)
for col_idx in range(num_columns):
if base_positions[col_idx] >= canvas_height // base_size and frame_idx < end_frame_start:
string_lengths[col_idx] = random.randint(10, 100)
base_positions[col_idx] = -string_lengths[col_idx]
visible_bases[col_idx] = 0
# Randomly choose whether to make each base lowercase
column_strings[col_idx] = [
random.choice([base.lower(), base]) if random.random() < lowercase_prob else base
for base in [random.choice(bases) for _ in range(string_lengths[col_idx])]
]
if string_lengths[col_idx] > 8:
random_start = random.randint(0, string_lengths[col_idx] - 8)
random_white_sequences[col_idx] = range(random_start, random_start + 8)
last_10_percent_start = max(0, int(string_lengths[col_idx] * 0.9))
for row_idx in range(min(visible_bases[col_idx], string_lengths[col_idx])):
base_position = base_positions[col_idx] + row_idx
if 0 <= base_position * base_size < canvas_height:
base = column_strings[col_idx][row_idx]
if base:
if row_idx == visible_bases[col_idx] - 1:
draw_base(draw, col_idx, base_position, base, font, fill_color=active_color)
elif row_idx >= last_10_percent_start:
alpha = 255 - int(((row_idx - last_10_percent_start) / (string_lengths[col_idx] - last_10_percent_start)) * 127)
draw_base(draw, col_idx, base_position, base, font, alpha=alpha, fill_color=base_colors[base.upper()])
elif random_white_sequences[col_idx] and row_idx in random_white_sequences[col_idx]:
draw_base(draw, col_idx, base_position, base, font, fill_color=active_color)
else:
draw_base(draw, col_idx, base_position, base, font, fill_color=base_colors[base.upper()])
if visible_bases[col_idx] < string_lengths[col_idx]:
visible_bases[col_idx] += 1
base_positions[col_idx] += 2
# Convert the image to numpy array to check unique pixel values
img_array = np.array(img)
if len(np.unique(img_array)) > 2: # Only append frames with more than two unique pixel values (avoid black frames)
# Enhance contrast and saturation
if enhance:
img = ImageEnhance.Brightness(img).enhance(enhance[0]) # Slightly increase brightness
img = ImageEnhance.Sharpness(img).enhance(enhance[1]) # Sharpen the image
img = ImageEnhance.Contrast(img).enhance(enhance[2]) # Enhance contrast
img = ImageEnhance.Color(img).enhance(enhance[3]) # Boost color saturation
frames.append(img)
for i in range(transition_frames):
alpha = i / float(transition_frames)
transition_frame = Image.blend(frames[-1], frames[0], alpha)
frames.append(transition_frame)
save_output(frames, output_path, fps, output_format)