TipDet

Video.1.mov

1. data pipeline

• step1: run generate_label_txt.py to get label.txt
• step2: run generate_key_OF_imgs.py to get Optical flow RGB images of keyframes for each video
• step3: run generate_sequence_inputs_label.py to get the SequenceInput_label_groupShuffle.txt
  Data of porcine insertion can be downloaded from here (x63m). Unzip and put them in the Dataset folder.
  Data of human insertion will be available once the paper is accepted.

2. training

run train_sequence_inputs.py

3. test

run test.py

4. References

1. https://github.com/Megvii-BaseDetection/YOLOX
2. https://github.com/bubbliiiing/yolox-pytorch
3. https://github.com/YuHengsss/YOLOV
4. https://github.com/tomrunia/OpticalFlow_Visualization
5. https://github.com/NVIDIA/DALI
6. https://github.com/opencv/opencv_contrib
7. https://github.com/lucidrains/vit-pytorch