UWSTF

This dataset was developed to support research into cost-effective emotion recognition using IoT-based low-resolution thermal imaging. It contains a novel collection of thermal facial expression images designed for training and evaluating machine learning models, particularly Vision Transformers (ViTs) and Convolutional Neural Networks (CNNs). The dataset enables comparative studies between traditional architectures and modern transformer-based approaches.

In addition to standalone experiments, the dataset has been utilised in hybrid training with the Kotani Thermal Facial Emotion (KTFE) dataset to investigate model generalisation. Results from these experiments demonstrate that ViT architectures not only train more efficiently but also achieve superior performance across multiple metrics compared to ResNet, both on this dataset and in cross-dataset evaluations with KTFE.

This resource offers researchers an accessible and affordable avenue for advancing thermal facial expression recognition. It underscores the growing potential of low-cost thermal imaging technologies, the competitive edge of transformer-based models in affective computing, and the importance of dataset diversity for robust emotion recognition systems.