End-to-End multimodal emotion recognition

About

I am a front-end engineer, due to some reasons need to do a front-end multi-modal emotion recognition system, but I am a python novice. Fortunately, I found @katerynaCh's open source library, which helped me a lot, click here&star it. While her open source library covers this code in enough detail, I just reorganized it so that the code can output result directly from the raw video file

Dependencies

Please use pip install -r requirements.txt to install the dependencies

Preparation&Usage

1. You need to download pth files EfficientFace_Trained_on_AffectNet7.pth.tar and lt_1head_moddrop_2.pth according to her library
2. Orgnize the file structure like this

   -root
       -models
           lt_1head_moddrop_2.pth
           ...
       -raw_data
           your_video.mp4
       EfficientFace_Trained_on_AffectNet7.pth.tar
       ...
       main.py
   

3. Due to I fixed the options. You can run python main.py directly.

Thanks Again

This work has been bothering me for a long time, because I don't know deep learning and python, and most of the open source code found online requires training, which my computer can't support me to complete. Thanks for the code and the trained model shared by @katerynaCh