Metadata-Version: 2.4
Name: license-plate-annotation-tool
Version: 0.1.6
Summary: An approachable browser tool for creating license-plate datasets from video.
Author: License Plate Annotation Tool contributors
License-Expression: MIT
Project-URL: Repository, https://github.com/hasan-farooq-hazen/LP-Annotation-Tool/tree/pypi-releases
Project-URL: Issues, https://github.com/hasan-farooq-hazen/LP-Annotation-Tool/issues
Keywords: license-plate,alpr,ocr,annotation,computer-vision
Classifier: Development Status :: 3 - Alpha
Classifier: Environment :: Web Environment
Classifier: Intended Audience :: End Users/Desktop
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Multimedia :: Graphics :: Capture :: Digital Camera
Requires-Python: <4.0,>=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: fast-alpr[onnx]<0.5.0,>=0.4.0
Requires-Dist: streamlit<2.0,>=1.40
Requires-Dist: opencv-python-headless<5.1,>=4.8
Dynamic: license-file

# License Plate Annotation Tool

[![PyPI](https://img.shields.io/pypi/v/license-plate-annotation-tool)](https://pypi.org/project/license-plate-annotation-tool/)
[![Python](https://img.shields.io/pypi/pyversions/license-plate-annotation-tool)](https://pypi.org/project/license-plate-annotation-tool/)

![License Plate Annotation Tool interface](https://raw.githubusercontent.com/hasan-farooq-hazen/LP-Annotation-Tool/pypi-releases/assets/license-plate-annotation-tool.png)

A browser-based tool for turning vehicle footage into clean, reviewable license-plate datasets.
It provides a guided workflow with useful defaults, so no computer-vision or command-line
expertise is required after installation.

Assembled and packaged by [hazen.ai](https://www.hazen.ai/), the tool combines existing
open-source packages in an approachable interface. It is powered by
[FastALPR](https://github.com/ankandrew/fast-alpr) and uses ONNX models for plate detection
and OCR.

## Quick start

Python 3.10 or newer is required.

1. Install the package:

   ```bash
   pip install license-plate-annotation-tool
   ```

   To update an existing installation:

   ```bash
   python -m pip install --upgrade license-plate-annotation-tool
   ```

2. Open the app:

   ```bash
   license-plate-annotation-tool
   ```

The app opens automatically in your browser. If the launcher is not available in the current
terminal, use:

```bash
python -m license_plate_annotation_tool
```

## What it does

The guided workflow handles the complete process from a single uploaded video:

1. **Create frames** — choose 5–20 FPS and optionally process only part of the video.
2. **Remove repeats** — group visually similar frames and retain the clearest examples.
3. **Extract plates** — detect and crop each plate, then use OCR text to name it when possible.

Each stage is also available independently when you only need one part of the workflow.

## Results

Results are provided as downloadable ZIP files. Depending on the selected workflow, they
include extracted frames, a reduced image set, or license-plate crops, together with:

- `manifest.csv` — a record of generated files and processing outcomes.
- `run_config.json` — the settings used for that run.

Unreadable plates are still retained, allowing them to be reviewed and labeled manually.

## Models

The default ONNX models are:

| Task | Default model | Provided through |
| --- | --- | --- |
| Plate detection | `yolo-v9-s-608-license-plate-end2end` | [Open Image Models](https://github.com/ankandrew/open-image-models) |
| Plate OCR | `cct-s-v2-global-model` | [Fast Plate OCR](https://github.com/ankandrew/fast-plate-ocr) |

Other models exposed by the installed FastALPR version can be selected in the app. Model files
are downloaded on first use, stored in the user's cache, and are not included in this package.
CPU inference is provided by ONNX Runtime.

## Privacy and storage

Uploads and generated files remain in a temporary directory for the active browser session.
Download any results you want to keep before closing or restarting the app.
