Implementación de Algoritmo de Detección y Segmentación de Rostros en Fotografías Mediante Procesamiento Digital de Señales.

In the present work the implementation of an algorithm of detection and segmentation of faces in photographs is proposed through the investigation of the state of the art in this area of computer vision; for an optimal choice of the algorithm to be implemented, a comparative analysis was performed between the Viola & Jones, MTCNN, YOLO detection algorithms. Additionally, the results obtained from the development of an algorithm based on the convolutional neural network (CNN) architecture are presented in this work. The comparison between the mentioned models will be carried out taking into account two specific case studies: photographs that contain only one face and photographs that contain several faces. For both cases a test set consisting of 15 and 25 images will be used respectively, with variations in the pose, ethnicity, makeup and occlusion. In addition, each photograph will be digitally processed using different normalization techniques such as: Histogram equalization, gamma correction, scale and rotation, in order to analyze the behavior of each algorithm in different settings. Finally, in images that contain only one face, the detection rate of the algorithm to be implemented will be calculated through the use of evaluation parameters such as: True positives, false negatives and false positives; while in multi-face images the detection rate will be determined by the F-Score parameter calculated in each image. Keywords: Face Detection, Segmentation, Viola & Jones, MTCNN, YOLO, CNN, Normalization.

• Glavni avtor: Orozco Calva, Soraya Paola (author)
• Format: bachelorThesis
• Jezik: spa
• Izdano: 2019
• Teme: <DETECCIÓN DE ROSTROS><SEGMENTACIÓN DE ROSTROS><ALGORITMOS><REDES NEURONALES CONVOLUCIONALES><YOLO>>
• Online dostop: http://dspace.unl.edu.ec/jspui/handle/123456789/22601