Accident Detection System

Real-time accident detection system using YOLOv8 and Gradio interface.

Setup

1. Clone the repository:

git https://github.com/Abood-devo/CCTV-Accident-Detection.git
cd CCTV-Accident-Detection

2. Create conda environment:

conda env create -f environment.yml
conda activate accident-detection

3. Download pre-trained weights:

mkdir -p models/weights
# Download weights to models/weights/best.pt

Usage

Start the Gradio interface:

python src/main.py

Access the interface at: http://localhost:7860

Training

1. Prepare your dataset following instructions in data/README.md
2. Configure training parameters in config/training_config.yaml
3. Run training notebook: notebooks/02_model_training.ipynb

Project Structure

• src/: Source code for inference and web interface
• data/: Dataset and data processing scripts
• models/: Model weights and configurations
• notebooks/: Jupyter notebooks for training and evaluation
• config/: Configuration files
• incidents/ : Detected accident clips and metadata