Emotion Recognition System
EmoTorch was built as a part of the FB AI Hakcthon 2020. We worked in a team of 3 to turn our idea into reality.
The motivation for this project came from the recommendation systems. Often we see products recommended based on the user’s search history, watch history , purchase history etc. These sort of systems depend on the past history and records. We with this project want to use real-time recommendation system. The system will use our model to predict the emotion of the user based on his facial expressions and will recommend products to suit the mood of the user.
The model is built entirely on the publicly available datasets. Since, the field Facial Emotion Recognition(FER) is not much developed yet we had very limited available datasets. Based on our research, we chose the FER dataset.
The data consists of 48x48 pixel grayscale images of faces. The faces have been automatically registered so that the face is more or less centered and occupies about the same amount of space in each image. The task is to categorize each face based on the emotion shown in the facial expression in to one of seven categories.
Class Label
The model is ready and predicts accurate emotions based on the image of a person’s face. Our aim to build this project is to merge it with Facebook’s product recommendation system.
The predicted labels will be sent to the recommendation system which will inturn predict the most likely products.
The front camera of a cell phone or laptop will capture the face of the person on consent while browsing Facebook Feed. Alternatively, a selfie can be fed. This image is sent to our model which predicts the emotion of the user. This emotion is sent to the recommendation system which predicts the products based on those emotions. These emotions can also be used to recommend things like Place to Visit, Food to Eat, Song to Listen and Movie to watch.
This will help in better product recommendation which in turn will increase the probability of customer buying or clicking on a product.
Check out on Devpost.