Enhancing Emotion Recognition in AI

A computer-vision project that improves AI emotion recognition by making the model more interpretable and less biased.

• Project page: https://lacygus.github.io/Emotion_Recognition
• Code: github.com/lacygus/Emotion_Recognition

What it does. Upgrades an EmotionROI classifier with ResNet-50 and concept models, and uses CLIP-Dissect to interpret what individual neurons respond to, surfacing and reducing biases in the learned representations while improving emotion-detection accuracy.

Python · PyTorch · ResNet-50 · CLIP · model interpretability