DERFormer: Driver Emotion Recognition Transformer via Dual Feature Enhancement for Driver Monitoring System

This repository is the official implementation of the paper DERFormer: Driver Emotion Recognition Transformer via Dual Feature Enhancement for Driver Monitoring System.

Prepare

Download pretrained weights (ir101.pt, pe.pt and vit.pt) from https://huggingface.co/TTJiang/DERFormer and place them into ./pretrained_weights

Run

python preprocess.py --cache-root Path/to/Cached/Data --mli-der-root Path/to/MLI-DER/Dataset --kmu-fed-root Path/to/KMU-FED/Dataset
python train.py --cache-root Path/to/Cached/Data/MLI-DER
python train.py --cache-root Path/to/Cached/Data/KMU-FED

License

MIT