Object tracking

Description

1. Yolov5 object detection
2. Simple Online and Realtime Tracking (SORT)
3. Deep Simple Online and Realtime Tracking (Deep SORT)
4. ByteTrack

Model library

2D Detection	Multi Object Tracking	KeyPoint Detection
Yolov5	1.Sort 2. Deep Sort 3. ByteTrack	1. TinyPose

Setup

conda create -n myenv python=3.9 -y

after then

pip install -r setup.txt

Run

"""
@weigths: file checkpoint
@source:  file input
@classes: filter classes
@type_mot: {null|sort|deep_sort} 
"""
export PYTHONPATH=.

python main.py --weights ./saved_models/yolov5s.pt \
               --source ./test_video/test.mp4 \
               --classes 0 32 --type_mot deep_sort

OR

1. Run object tracking

# Run object detection
sh scripts/object_detect.sh

2. Run object detection + tracking (yolov5 + sort)

# Run object detection + tracking (yolov5 + sort)
sh scripts/object_sort.sh

3. Run object detection + tracking (yolov5 + deep sort)

# Run object detection + tracking (yolov5 + deep_sort)
sh scripts/object_deep_sort.sh

References

1. ultralytics/yolov5: YOLOv5 🚀 in PyTorch > ONNX > CoreML > TFLite (github.com)
2. https://github.com/abewley/sort
3. https://github.com/ZQPei/deep_sort_pytorch

Conclusion

This documentation has demonstrated how to use related module. Before actually start working on anything, please read the whole document first. If you need any clarifications, please contact me. Thanks for reading and good luck on improving the model.

Happy Coding