Here is the PyTorch implementation of the EmoHEAL model. EmoHEAL is a lightweight deep learning architecture that uses wrist sensor data and achieves 72.35% accuracy on the K-EmoCon dataset, making it suitable for IoT applications in smartwatches.
- TCN+CA-SA+GRU Fusion
- TCN+GRU Fusion
- TCN+xLSTM
- TCN+MHA
- TCN+Transformer Encoder
- ResNet + GRU Fusion
You can request the dataset from K-Emocon. Place the dataset in the dataset
folder.
Dependencies can be installed using the following command:
conda env create -f EmoHEAL.yml