Deep Convolution Neural Network for Facial Emotional Recognition

Based on my on-going progress on my Master's thesis on Deep Convolution Neural Network for Facial Emotion Recognition. The network is based on VGG-16 trained on 8 labeled emotions using more than 4,000 pictures. Transfer learning technique is applied on the existing VGG-16 network. All work is being made through MATLAB.

Initial Sketch

Applications

There are many interesting use case scenarios for this project. It can be used as a happiness meter for commercial entities to rate the service provided using customer's facial expression rather than presenting a customer with 3 buttons which stand for happy, neutral, not happy. Also this project can be used within a robot, or a machine for similar purposes as well.

Privacy Concerns

The network only saves customer's picture for the purpose of classifying his facial expression, the image saved is temporary and is over written by the next picture captured. The project does not require further training on new images as the training has already been done on over 4,000 pictures. In addition, to run the network, no internet connection is required.

Project Presentation

Final project documentation and files can be found here.