README.md

---
title: Pyronear Wildfire Detection
emoji: 🏆
colorFrom: blue
colorTo: pink
sdk: streamlit
python_version: 3.11
app_file: app.py
pinned: false
license: apache-2.0
---

# MP4 to 8 Frames + Wildfire Detection for Pyronear

Upload an MP4, extract evenly spaced frames, run wildfire detection on each,
and display the main detections (one image per main detection).

## Requirements
- Python 3.9+
- Packages listed in `requirements.txt`

## Install
```bash
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
```

## Run
```bash
streamlit run app.py
```

Streamlit will print a local URL (for example, `http://localhost:8501`). Open it
in your browser and upload an MP4. Detection starts automatically after upload.

## Docker Compose + Make
Run with:
```bash
make run
```

Other commands:
```bash
make logs
make stop
make down
```

Then open `http://127.0.0.1:7860` in your browser.

## Notes
- The first run downloads the wildfire detection model from Hugging Face.
- `ffmpeg`/`ffprobe` are required for frame extraction.
- OpenCV is used for motion features and image processing.