Skip to content

GaitGAN: Invariant Gait Feature Extraction Using Generative Adversarial Networks

Notifications You must be signed in to change notification settings

xuehy/pytorch-GaitGAN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GaitGAN

A pytorch implementation of GaitGAN: Invariant Gait Feature Extraction Using Generative Adversarial Networks.

Yu, Shiqi, et al. "Gaitgan: invariant gait feature extraction using generative adversarial networks." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 2017.

Dependency

Training

To train the model, put the CASIA-B dataset silhoutte data under repository Then goto src dir and run

python3 train.py

The model will be saved into the execution dir every 500 iterations. YOu can change the interval in train.py.

Monitor the performance

  • Install visdom.
  • Start the visdom server with python3 -m visdom.server 5274 or any port you like (change the port in train.py and test.py)
  • Open this URL in your browser: http://localhost:5274 You will see the loss curve as well as the image examples.

After 19k iterations, the results(every 3x1 block shows the generated side view, ground truth side view and the input view GEI in order):

19

the loss curve is:

loss19k

Testing

  • goto src dir and run python3 test.py
  • Open this URL in your browser: http://localhost:5274 You will see the results on the test set.

After 19k iterations, some of the results: test19k

Recognition

The codes for recognition are also provided.

The dataset setting is identical to the paper, while we only test ProbeMN here.

  • Goto src dir and mkdir transformed_28500
  • run python3 generate.py
  • run python3 knn_class.py, you'll get the average accuracy with KNN(k=1) on ProbeMN.
  • run python3 knn_class_per_angle.py, you'll get the results for different Gallery views and Probe views.

About

GaitGAN: Invariant Gait Feature Extraction Using Generative Adversarial Networks

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages