This project demonstrates the use of Deep Learning to detect emotion (sad, angry, happy etc) from the images of faces.
Challenges in Representation Learning: Facial Expression Recognition Challenge
- Shallow fully connected network
- Mini VGG16
- Mini GoogLeNet
- Shallow CNN with progressively increasing channels
- Mini VGG16 with LSUV
The combination of VGG16, SGD, a bit of data augmentation, and Label Smoothing yielded the best generalization.
Experimentation report available here: https://app.wandb.ai/sayakpaul/emotion-detection
I happily welcome any feedback :)