DETEKSI WAJAH MANUSIA BERBASIS ONE STAGE DETECTOR MENGGUNAKAN METODE YOU ONLY LOOK ONCE (YOLO)

Abstract

The purpose of face detection is to find the location of the face in an image; hence it can be utilized for further applications such as face recognition, finding specific faces in a video, and others. We present Algorithm You Only Look Once based on the one-stage detector approach. This algorithm divided an image into grids of a specific size, where each grid or cell represents the candidate location of the target object. The implemented architecture of the Convolutional Neural Network of this algorithm is to classify the candidate object within each grid into a face or non-face class and justify the face's location. We used the trainable VGG-16 model for the convolutional layers and trained the fully connected layers with the appropriate target label. The experiments are conducted using the WIDER Face dataset with various face objects in each image. As a result, we achieved the highest precision, recall, and f1-score are 0.253, 0.247, and 0.25