Skip to content

Spatial Temporal Graph Convolutional Networks for Emotion Perception from Gaits

License

Notifications You must be signed in to change notification settings

UttaranB127/STEP

Folders and files

NameName
Last commit message
Last commit date

Latest commit

797b53f · Nov 8, 2021

History

55 Commits
Aug 23, 2020
Nov 8, 2021
Jun 29, 2020
Aug 23, 2020
Nov 8, 2021
Jun 29, 2020
Aug 23, 2020
Jul 3, 2020
Feb 3, 2021
Oct 10, 2021

Repository files navigation

This is the official implementation of the paper STEP: Spatial Temporal Graph Convolutional Networks for Emotion Perception from Gaits. Please use the following citation if you find our work uesful:

@inproceedings{bhattacharya2020step,
author = {Bhattacharya, Uttaran and Mittal, Trisha and Chandra, Rohan and Randhavane, Tanmay and Bera, Aniket and Manocha, Dinesh},
title = {STEP: Spatial Temporal Graph Convolutional Networks for Emotion Perception from Gaits},
year = {2020},
publisher = {AAAI Press},
booktitle = {Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelligence},
pages = {1342–1350},
numpages = {9},
series = {AAAI’20}
}

We have also released the Emotion-Gait dataset with this code, which is available for download here: https://go.umd.edu/emotion-gait.

  1. generator_cvae is the generator.

  2. classifier_stgcn_real_only is the baseline classifier using only the real 342 gaits.

  3. classifier_stgcn_real_and_synth is the baseline classifier using both real 342 and N synthetic gaits.

  4. clasifier_hybrid is the hybrid classifier using both deep and physiologically-motivated features.

  5. compute_aff_features consists of the set of scripts to compute the affective features from 16-joint pose sequences. Calling main.py with the correct data path computes the features, and save them in the affectiveFeatures<f_type>.h5 file, where f_type is the desired type of features:

    • '' original data (default)
    • 4DCVAEGCN data generated by the CVAE.