• torch==1.8.1
• timm==0.4.12
• opencv-python==4.8.0.76
• sscipy==1.10.1
• yacs==0.1.8
• loguru==0.7.2
• pandas==2.0.3
• hdf5storage==0.1.19
• matplotlib==3.4.3
• numpy==1.21.2

cd S3_SAM_CSD
pip install -e .

• downloaod ordinary train video
  • Baidu link : https://pan.baidu.com/s/1PrRw1Pjmh-Iu8pssVuJW3Q?pwd=12rl
    • pwd：12rl
• put the video in directory [project]/data

  • The directory structure is as follows:

    [project]
    ├─ data
    │    └─ csd.mp4
    ├─ S0_OpticalFlow
    ├─ S1_Classification
    ├─ S2_BuildPrompts
    ├─ S3_SAM_CSD
    ├─ img2video.py
    ├─ requirements.txt
    ├─ run.sh
    ├─ README.md
    └─ video2img.py
    

python video2img.py

./run.sh

• FlowFormer
• SAM

The model is licensed under the Apache 2.0 license.