FACE RECOGNITION THROUGH CONVOLUTIONAL NEURAL NETWORK

Face recognition involves identifying a face and checking whether the face already exists in the system or not. The proposed work is focused on capturing face, predicting emotion, finding out location and storing date and time when an image is captured. This involves usage of a deep learning technique by the name convolutional neural network. A CNN is built to process the image and predict an emotion from the face. A data file is created comprising the stated details for the further assessment. An emotion plays an important role in finding out how a person is acting when doing a particular thing. It sometimes conveys the intention of a person. Date and time makes us aware of the time when an image was previously captured. The location attribute is the location of the internet service provider of an area. Python programming language is utilized to implement the idea of constructing a face recognition system. This comes with a set of useful libraries to ease our task. It is observed that these face recognition systems take a lot of time in training and the proposed work focuses on reducing it to a certain extent and to recognize faces thus paving a way for an improved face recognition rate. Fer2013 dataset is used to train the CNN model and it involves training it with a face and an associated emotion. The shortcomings of the recognition systems which include difficulty in finding faces, spending a considerable amount of time in training and testing are considered to be improved. Some of the face recognition systems built using R-CNN and FRR-CNN exhibited recognition rates less than 80%. The proposed work intend to improve the face recognition rate a bit more than the existing ones.