POLIBITS, Vol. 60, pp. 67-72, 2019.
Abstract: Convolutional neural network, facial recognition, artificial vision, few-shot learning
Keywords: The paper presents a feature-based face recognition method. The method can be explained in two separated processes: A pretrained CNN-Based face detector looks for faces in images and return the locations and features of the found faces, this face detector will be used to train the models for the classifiers and then will be used to find unknown faces in new images. The used classifiers are: K-Nearest Neighbors, Gaussian Naive Bayes and Support Vector Machines. Each model will be trained with a different quantity of training examples in order to obtain the best version of the method. When the models are ready, each classifier will try to classify the faces with the previously trained models. The accuracy of each classifier in few-shot face recognition tasks will be measured in Recognition Rate and F1 Score, a comparative table of the results is presented. This paper has the goal to show the high accuracy achieved by this method in datasets with several individuals but few examples of training.
PDF: Facial Recognition using Convolutional Neural Networks and Supervised Few-Shot Learning
PDF: Facial Recognition using Convolutional Neural Networks and Supervised Few-Shot Learning
https://doi.org/10.17562/PB-60-8
Table of contents of POLIBITS 60