An occluded pedestrian detection project using improved YOLOv5s model
- Improved YOLOv5s model to enhance occluded object detection capabilities by embedding SE attention mechanisms and replacing SPP with SPP-Fast modules in backbone and adding specialized detection head for small objects
- Constructed and trained improved model in PyTorch based on dataset of 2,000+ images with occluded pedestrians
- Data Processing: Python - Pandas, Numpy
- Modeling: Python - PyTorch
Pedestrian Detection
├── data/ # trajectory dataset
├── models/
├── runs/
├── utils/
├── venv/
├── src/
| ├── detect.py # detection
| ├── export.py # export model to other formats
| ├── train.py # train
| └── val.py # validate
├── requirements.txt
├── README.md
└── LICENSE
Feel free to open issues and pull requests to contribute!