Inspiration

Facial detection is something that is more difficult for some people than for others, and extremely difficult for computers. Simplifying this process seemed like a difficult and challenging task.

What it does

A machine learning model using Torch that identifies one of seven possible facial expressions present in a 48x48 image. 0:'Angry', 1:'Disgust', 2:'Fear', 3:'Happy', 4:'Sad', 5:'Surprise', 6:'Neutral'

How we built it

A hefty amount of preprocessing using PCA by SVD and the original normalized data.

Challenges we ran into

Originally wanted to develop an unsupervised model, it was increasingly difficult to find the type of distribution function responsible for expression detection.

Accomplishments that we're proud of

having simplified the input space by a factor of more than one hundred, I was able to get fairly accurate results.

What we learned

How to implement highly customized torch models.

What's next for Emotion Recognition

Optimization of the network. Increasing recognition of unseen samples.

Built With

• numpy
• pandas
• pytorch