Here is a video of me explaining the code and showing its working: https://youtu.be/wLbgkwqNlWc
The python libraries required to run the code is there in the requirements.txt file.
The emotion recognition model accuracy is about 63%. You can use my pretrained models with weights or build a new model from scratch.
The main.py file has the main script. The paths in the file are hard coded so make sure you edit them before running the script.
You might also need ffmped to convert the video to audio file. Here is the link https://ffmpeg.zeranoe.com/builds/ Download and place in C: drive.
You will need to install Nvidia CUDA 11.0+
The design and training of the emotion recognition model have been is in the "Emotion_Recognition_Model.ipynb" file. The model structure and weights were then saved.
The weights and model were then loaded in main.py file before using them on a video.
feel free to email me at: [email protected] :)