Multimodal-Video-Emotion-Recognition-Pytorch

This repository contains code for training different uni-modal and multi-modal emotion recognition models written in Pytorch. This work was done as an assignment for the Affective Computing course (Spring'20) at IIIT Delhi. The file Report.pdf provides more details.

Final Architectures

Feature Extraction

Hybrid Fusion model

MTL model

Data

All the required pickle files can be found here: https://drive.google.com/drive/folders/1x4rlJwvhYH_kxakuiX8s6iwQvnN_yfQa?usp=sharing.

Results

Baseline results

Final results