Metadata-Version: 2.4
Name: visimac
Version: 1.0.1
Summary: VisiMac - High-Performance Deep Learning, Computer Vision & Dataset Preparation Engine
Requires-Python: >=3.9
Description-Content-Type: text/markdown
Requires-Dist: PyQt6
Requires-Dist: opencv-python
Requires-Dist: numpy
Requires-Dist: torch
Requires-Dist: pillow
Requires-Dist: pandas
Requires-Dist: matplotlib
Requires-Dist: pyyaml

<div align="center">

# ⚡ VisiMac Studio

### **High-Performance Computer Vision & Dataset Preparation Engine**

[![PyPI Version](https://img.shields.io/pypi/v/visimac.svg?style=flat-record&color=6366F1)](https://pypi.org/project/visimac/)
[![Python Version](https://img.shields.io/badge/python-3.9%20%7C%203.10%20%7C%203.11%20%7C%203.12%20%7C%203.13%20%7C%203.14%20%7C%203.15-blue)](https://www.python.org/)
[![Platform](https://img.shields.io/badge/platform-Windows%20%7C%20macOS%20%7C%20Linux-lightgrey)](#-cross-platform-support)

---

**VisiMac Studio** is an all-in-one workstation designed for computer vision engineers, AI researchers, and data scientists. It bridges the gap between raw data collection and production model deployment by integrating interactive dataset annotation, multi-unit image batch processing, grid tile splitting, synthetic data augmentation, AutoML hyperparameter optimization, YOLO training/metrics tracking, and model export.

</div>

---

## 🌟 Key Features

| Module | Features & Capabilities |
| :--- | :--- |
| **🖼️ Interactive Annotator** | Multi-shape drawing (Polygon, Rectangle, Circle, Pentagon, Move/Select), Perspective Warp, Fixed/Custom Rotations, Point Crop, Image Sequence Frame-Copying, and AI Auto-Labeling (`.pt` weights). |
| **📐 Batch Scaling & Compression** | Multi-unit conversion (`px`, `cm`, `mm`, `in`, `pt`), DPI target adjustments, aspect-ratio lock, and **exact target file size binary-search quality compression** (in KB). |
| **✂️ Tile Splitter Engine** | Dynamic row/column tile splitting, structured subfolder grouping (`part_1/`, `part_2/`), and zero-padded naming schemas. |
| **🎨 Synthetic Augmentation** | Real-time live simulation of YOLO augmentations (Mosaic, Mixup, HSV jitter, Shearing, Translation, Crop Fraction, Random Erasing) and batch synthetic image generation. |
| **🧬 YOLO Training & AutoML** | Multi-backbone architecture support, genetic algorithm hyperparameter tuning (AutoML), dataset quality health inspection (unannotated/orphaned check), real-time loss graph rendering, and evaluation breakdown. |
| **⚡ Inference & Production Export** | Batch directory inference with confidence thresholding and one-click model exporting to production formats (`ONNX`, `TensorRT/Engine`, `OpenVINO`, `TorchScript`, `CoreML`). |

---

## 🛠️ Installation

### Standard Installation via PyPI

Install the pre-compiled binary package directly using `pip`:

```bash
pip install visimac
```

Launch the workstation GUI directly from your terminal:

```bash
visimac
```

---

## 🚀 Workstation Guide

### 1. Interactive Workspace & Annotator
* **Load Directory:** Load image folders (`.jpg`, `.png`, `.bmp`, `.webp`) along with any existing `classes.txt` or YOLO format `.txt` label files.
* **Transformation Tools:** Perform non-destructive image rotation (fixed 90° CW/CCW or custom precision angles), point cropping, and 4-point perspective warping directly on the canvas.
* **AI Assist & Frame Copying:** Connect a pre-trained PyTorch YOLO model (`.pt`) to automatically label target frames based on confidence thresholds, or copy shapes across sequential camera frames.
* **Dataset QA Health Inspector:** Identify unannotated images and orphaned label files before running expensive model trainings.

### 2. Batch Resizer & Compression Engine
* **Dimension Controls:** Scale images by precise pixel counts or physical units (`cm`, `mm`, `in`, `pt`) tied to custom target DPIs.
* **Size Constraint Search:** Specify a target maximum file size in kilobytes (e.g., `500 KB`). The built-in binary search algorithm calculates the optimal compression quality factor per image without destroying aspect ratios or visual fidelity.

### 3. Tile Splitter Engine
* Decompose large high-resolution satellite, medical, or industrial images into structured image grids.
* Select target slice counts (e.g., `4` parts $\rightarrow$ `2 Rows x 2 Cols`).
* Optionally organize tiles into dedicated indexed subfolders (`part_1/`, `part_2/`) for parallel model ingestion.

### 4. Synthetic Augmentation Studio
* Adjust HSV saturation/hue offsets, translation shifts, rotation ranges, mosaic probabilities, and cutout parameters.
* Click **Simulate Augmentation Canvas** to view how YOLO will transform your datasets during training.
* Batch-generate synthetic variants directly into mapped project directories.

### 5. Training, AutoML & Inference
* **AutoML Optimization:** Run genetic algorithms across mini-epochs to automatically determine optimal learning rates (`lr0`, `lrf`) and augmentation parameters.
* **Live Loss Analytics:** Monitor real-time training progress via live-rendered loss curves and metric breakdowns (`Precision`, `Recall`, `mAP50`, `mAP50-95`).
* **One-Click Export:** Convert standard `.pt` model weights into edge-ready deployment binaries (`ONNX`, `TensorRT/Engine`, `OpenVINO`, `TorchScript`, `CoreML`).


---

## 💻 Cross-Platform Support

| Operating System | Binary Extension Type | Architecture |
| :--- | :--- | :--- |
| **Windows** | `.pyd` | `x86_64` / `AMD64` |
| **Linux** | `.so` | `x86_64` |
| **macOS** | `.so` | `x86_64` / `arm64` (Apple Silicon) |

---

<div align="center">
  <sub>Built for computer vision workflows.</sub>
</div>
