Abstract
Face recognition is widely used in many practical applications. It also receives great attention from the scientific community in computer vision. Although face detection and facial recognition algorithms using convolutional neural networks highly regarded for their accuracy, they still face the following major challenges: the dataset must be large enough, a large amount of computational resources are required for the training and testing phases, long training times, many network configurations need to be tested. Therefore, in this thesis, we are attention to using scattering network for facial recognition. The thesis provides a general process of the face recognition: generating sample data from original images, image preprocessing, feature extraction using ScatNet, using SVM to classify faces, training the network model with sample data, face recognition. In addition, we develop an employee attendance application according to the above process. Experiments on feature extraction speed, training time, recognition speed and accuracy on sample data with different parameters of the network model are also introduced in the thesis. Finally, we summarize the results and give directions for further development.