Skip to content

A test version of Yolov5 in PyTorch for object detection

License

Notifications You must be signed in to change notification settings

shangjie-li/yolov5-test

Repository files navigation

yolov5-test

A test version of Yolov5 in PyTorch for object detection

Acknowledgement

Dataset

  • Check COCO 2017 dataset
    python dataset_player.py
    python dataset_player.py --training
    
  • Check KITTI dataset
    python dataset_player.py --data=data/kitti.yaml
    python dataset_player.py --data=data/kitti.yaml --training
    
  • Check SEUMM HQ LWIR dataset
    python dataset_player.py --data=data/seumm_hq_lwir.yaml
    python dataset_player.py --data=data/seumm_hq_lwir.yaml --training
    
  • Check SEUMM LWIR dataset
    python dataset_player.py --data=data/seumm_lwir.yaml
    python dataset_player.py --data=data/seumm_lwir.yaml --training
    

Training

  • Train on COCO 2017 dataset
    python train.py
    
  • Train on KITTI dataset
    python train.py --data=data/kitti.yaml
    
  • Train on SEUMM HQ LWIR dataset
    python train.py --data=data/seumm_hq_lwir.yaml
    
  • Train on SEUMM LWIR dataset
    python train.py --data=data/seumm_lwir.yaml
    

Evaluation

  • Evaluate on COCO 2017 dataset
    python test.py --weights=weights/coco/yolov5s_100ep.pt
    
  • Evaluate on KITTI dataset
    python test.py --data=data/kitti.yaml --weights=weights/kitti/yolov5s_100ep.pt
    
  • Evaluate on SEUMM HQ LWIR dataset
    python test.py --data=data/seumm_hq_lwir.yaml --weights=weights/seumm_hq_lwir/yolov5s_100ep.pt
    
  • Evaluate on SEUMM LWIR dataset
    python test.py --data=data/seumm_lwir.yaml --weights=weights/seumm_lwir/yolov5s_100ep.pt
    
  • The result should be
Model Dataset Epoch val [email protected] val [email protected]:.95
yolov5s COCO 100 0.510 0.314
yolov5s KITTI 100 0.647 0.388
yolov5s (pretrained) KITTI 100 0.679 0.430
yolov5s SEUMM-HQ-L 100 0.952 0.613
yolov5s (pretrained) SEUMM-HQ-L 100 0.966 0.650
yolov5s SEUMM-L 100 0.886 0.564
yolov5s (pretrained) SEUMM-L 100 0.906 0.596

Demo

  • Run a demo with COCO 2017 model
    python detect.py --weights=weights/coco/yolov5s_100ep.pt --source=my_image.jpg --view-img
    python detect.py --weights=weights/coco/yolov5s_100ep.pt --source=test_images
    
  • Run a demo with KITTI model
    python detect.py --weights=weights/kitti/yolov5s_100ep.pt --source=my_image.jpg --view-img
    python detect.py --weights=weights/kitti/yolov5s_100ep.pt --source=test_images
    
  • Run a demo with SEUMM HQ LWIR model
    python detect.py --weights=weights/seumm_hq_lwir/yolov5s_100ep.pt --source=my_image.jpg --view-img
    python detect.py --weights=weights/seumm_hq_lwir/yolov5s_100ep.pt --source=test_images
    
  • Run a demo with SEUMM LWIR model
    python detect.py --weights=weights/seumm_lwir/yolov5s_100ep.pt --source=my_image.jpg --view-img
    python detect.py --weights=weights/seumm_lwir/yolov5s_100ep.pt --source=test_images
    

About

A test version of Yolov5 in PyTorch for object detection

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages