Skip to content

Latest commit

 

History

History

gesture

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
download_data.py
download_data.py
 
 
loader.py
loader.py
 
 
network.py
network.py
 
 
readme.md
readme.md
 
 
video_classification.py
video_classification.py
 
 

readme.md

Surgical Gesture and Skill

alt text

Data

This is a tutorial that uses the public JIGSAWS dataset for example applications of gesture recognition and skill assessment.

The JIGSAWS dataset can be downloaded at the linked site. The dataset will be used in this tutorial is the "video" folder and the meta files in each of the "Knot_Tying", "Needle_Passing" and "Suturing" tasks. Once downloaded, create a directory named "data", extract and copy the folder structure into that directory. The resulting directory structure should be:

gesture/data/
    -> Knot_Tying/
        -> meta_file_Knot_Tying.txt
        -> readme.text
        -> video/
            -> Knot_Tying_B001_capture1.avi
            -> Knot_Tying_B001_capture2.avi
            -> Knot_Tying_B002_capture1.avi
            ...
    -> Needle_Passing/
        -> meta_file_Needle_Passing.txt
        -> readme.text
        -> video/
            -> Needle_Passing_B001_capture1.avi
            -> Needle_Passing_B001_capture2.avi
            -> Needle_Passing_B002_capture1.avi
            ...
    -> Suturing/
        -> meta_file_Suturing.txt
        -> readme.text
        -> video/
            -> Suturing_B001_capture1.avi
            -> Suturing_B001_capture2.avi
            -> Suturing_B002_capture1.avi
            ...

Alternatively, run the provided data downloading script to download and extract the data to the required format:

python download_data.py

Once data are downloaded, run the video_classification.py script in the mphy0043 conda env:

conda activate mphy0043
python video_classification.py