Skip to content

Latest commit

 

History

History
31 lines (24 loc) · 1.07 KB

README.md

File metadata and controls

31 lines (24 loc) · 1.07 KB
Raw

PPRL-VGAN

Implementation of "VGAN-Based Image Representation Learning for Privacy-Preserving Facial Expression Recognition".

Prerequisites

  1. Tensorflow 1.0.1 or above
  2. Keras 2.0.2
  3. FERG facial expression database
  4. MUG facial expression database

Usage

First download both datasets from the offical sources. Then, reduce the video resolution to 64x64x3, and save the data into ".mat" or ".hdf5" files. In our experiments, we randomly chose 85% of the frames in each video for training and the rest frames for testing. You can decide your own training/testing ratio.

Training

python train.py

Evaluation of identification (privacy) and expression recognition (utility) performance

python Privacy_vs_Utility_check.py

Image processing applications (e.g., expression morphing)

python img_applications.py

Project webpage

http://vip.bu.edu/projects/vsns/privacy-smartroom/facial-expression-vgan/