YOLOv11n Object Detection on Video

This project performs real-time object detection on a video using the YOLOv11n model with GPU acceleration. The processed video with bounding boxes and confidence scores is saved as an output file.

Features

• Detect objects in a video using YOLOv11n.
• GPU acceleration with PyTorch.
• Draw bounding boxes with labels and confidence scores.
• Save processed video.
• Optional real-time display of processed frames.

Requirements

• Python 3.12
• GPU with CUDA support for faster inference.
• Packages listed in requirements.txt.

Setup

1. Clone the repository:

git clone https://github.com/ayhmdalila/traffic-detection
cd traffic-detection

2. Install dependencies:

pip install -r requirements.txt

Usage

Run the main script:

python main.py

Run Docker container:

docker run --gpus all -v $(pwd):/app yolo-video-detector

• Make sure to mount the project directory to /app in the container.
• Ensure your GPU drivers and Docker NVIDIA runtime are properly set up.

Configuration

• Frame size: Currently set to 640x360.
• Frame skipping: Can skip frames by adjusting frame_skip.
• YOLO confidence threshold: Set via conf parameter (default 0.4).
• IOU threshold: Set via iou parameter (default 0.5).

Notes

• GPU is required for optimal performance.
• For CPU-only mode, remove .to("cuda") in main.py.
• Supports .mp4, .mkv, and other common video formats.

License

Open source