Facial Emotion Recognition using Convolutional Neural Networks

In this modern generation, where we are frequently facing factors which build on our high levels of stress, the human race is facing a big problem of deteriorating mental health. Moreover, in the current scenario of Covid-19, this issue has worsened because of fewer social interactions than ever before, increase in screen time and even lack of physical work. Estimating facial emotion has forever been a simple task for people, however, accomplishing a similar perfection with a computer is very difficult. With the new progression in computer vision and AI, it is feasible to distinguish feelings from pictures and videos. When we decided on the idea of research, our goal was to assist people having issues related to mental health and make them come out of the pain they are suffering from. The main idea behind our work was the motto that it acts as effective research, not only a comparative analysis. The execution was done using a web-app. In the emotion recognition and detection component, which is the main component of our research, Harr Cascades was utilized to find and detect the face of an individual in front of the webcam. Haar Cascades has been used for the detection of the facial region captured by the webcam. It works as follows: instead of applying all the six thousand features present in a frame, it groups the features into different stages and applies them one after another and not all the six thousand features at once. The next component would be the Deep Learning algorithm developed using CNN. OpenCV has been used for the entire development. Tensor Flow has been utilized to train the model that has been developed. The model works by running frame by frame in order to detect the emotion. The facial region that has been detected through the Haar Cascade Classifier is passed as input to the Convolutional Neural Network model that has been developed by us. The emotion having maximum softmax scores is given as the result on live webcam feed itself.