Robust Pretreatment Strategy for Gait Recognition, view the paper here.
pip install -r requirements.txt
Download URL: CASIA-B.
- Please adjust the raw video file of the CAISA-B dataset to the following format:
CASIA-B-video
001-bg-01-000.avi
......
124-nm-06-180.avi
- Extracting silhouettes from CAISA-B-Video using GaitPretreatment. (including RVM, Clean, and Augment)
python CASIA-B.py
Download URL: Outdoor-Gait (Baidu Yun with extract code (tjw0) OR Google Drive).
- Please adjust the raw RGB image file of the OutdoorGait dataset to the following format:
OutdoorGait-img
images
001
scene1_bg_L_090_1
001.jpg
......
scene1_cl_L_090_1
002
...
- Extracting silhouettes from OutdoorGait-img using GaitPretreatment. (including RVM, Clean, and Augment)
python OutdoorGait.py
img_size
: Image resizing size. Default 64.
clean
: Clean up the unqualified silhouettes using a silhouette classification model. Default True.
augment
: Flip the cleaned silhouette sequence horizontally for data augmentation. Default True.
frame_resize_threshold
: When using RVM preprocessing, in order to speed up the silhouette extraction, images larger than the width of threshold pixels will be scaled to threshold pixels. Default 800 pixels.
pixel_threshold
: The threshold for finding the upper and lower boundaries of a person when cutting images. (the sum of row pixels is greater than or equal to the threshold pixel). Default 800 pixels.
@inproceedings{han2022gaitpretreatment,
title={GaitPretreatment: Robust Pretreatment Strategy for Gait Recognition},
author={Han, Yuanyuan and Wang, Zhong and Han, Xin and Fan, Xiaoya},
booktitle={2022 International Conference on Communications, Computing, Cybersecurity, and Informatics (CCCI)},
pages={1-6},
year={2022},
organization={IEEE}
}
Note: GaitPretreatment is only used for academic purposes.