One-Shot Learning Menggunakan Siamese Neural Network Untuk Pendeteksian Wajah

Abstract

One-Shot Learning using Siamese Neural Network is a method used for facial recognition, this method implements a deep learning algorithm which allows facial detection using one single image. This research aims to extract features from the images and calculate the dissimilarity of each compared images. This research uses Convolutional Neural Network (CNN) to perform feature extraction, where the images will be divided into two pairs and fed into two different CNN networks, then calculate the Eucledian Distance to measure similarity, calculate Contrastive Loss, and train Siamese Network to produce accurate results. The results of this research show that the algorithm can produce accurate results with high mean dissimilarity value for non-similar images and low mean dissimilarity value for similar images.