Resta Adityatama, Anggyi Trisnawan Putra

Image classification of Human Face Shapes Using Convolutional Neural Network Xception Architecture with Transfer Learning

Introduction

Image classification of human face shapes using convolutional neural network xception architecture with transfer learning. Classify human face shapes using CNN Xception architecture with transfer learning. Achieved 81.125% validation and 85.1% new data accuracy, enhancing facial recognition systems.

Abstract

Abstract. The development of information technology in facial recognition is influenced by a faster and more accurate authentication system. This allows the computer system to identify a person's face. Purpose: Similar to fingerprints and the retina of the human eye, each person's face has a different shape and contour. Since it is known that the human face provides a lot of information, as well as topics that attract attention make it studied intensively. Methods/Study design/approach: Several studies examining information from human faces are facial recognition. One of the approaches used to recognize facial imagery is through the use of a Convolutional Neural Network (CNN). CNN is a method in the field of Deep Learning that can be used to recognize and classify objects in digital images. In this study, the method used to implement facial image classification is the Xception architecture CNN algorithm with a transfer learning approach. Result/Findings: The dataset used in this study was obtained from Kaggle, namely the Face Shape Dataset which contains 5000 data. After testing, an accuracy rate of 96.2% was obtained in the training process and 81.125% in the validation process. This study also uses new data to test the model that has been made, and the results show an accuracy rate of 85.1% in classifying facial imagery. Novelty/Originality/Value: Therefore, it can be said that the model created in this study has the ability to classify images of facial shapes Human Face Shapes Using Convolutional Neural Network Xception Architecture with Transfer Learning.

Review

The manuscript, "Image classification of Human Face Shapes Using Convolutional Neural Network Xception Architecture with Transfer Learning," addresses the pertinent task of classifying human face shapes using deep learning techniques. The authors propose a methodology based on the Xception Convolutional Neural Network architecture, enhanced with a transfer learning approach. Utilizing a Kaggle dataset of 5000 facial images, the study reports promising accuracy rates, achieving 96.2% during training, 81.125% during validation, and 85.1% when tested on new data. The topic is highly relevant given the ongoing advancements in facial recognition and the potential information embedded in facial contours. A significant strength of this work lies in its application of a contemporary and powerful deep learning model, Xception, combined with transfer learning. This is a well-established and effective strategy for image classification, particularly when working with potentially limited or domain-specific datasets. The abstract clearly outlines the chosen method and reports specific quantitative results, which is a positive aspect for an initial assessment. The use of a publicly available dataset like Kaggle's Face Shape Dataset also enhances the potential for reproducibility, although further details on the dataset's specific characteristics (e.g., number of classes, class distribution) would be beneficial. The reported accuracy on new data suggests a degree of generalization capability for the developed model. However, the abstract presents several areas that require more clarity and deeper elaboration to fully assess the work's contribution. Firstly, the "Purpose" and "Novelty/Originality/Value" sections are rather generic; the purpose could be more narrowly defined to focus on a specific research gap, and the novelty statement largely reiterates the approach without clearly articulating what makes this study a distinct or significant advancement beyond existing literature. The notable discrepancy between the high training accuracy (96.2%) and the lower validation accuracy (81.125%) suggests potential overfitting, which needs to be thoroughly discussed, analyzed, and potentially mitigated. Furthermore, the accuracy on "new data" (85.1%) being higher than the validation accuracy is unusual and warrants a detailed explanation of the dataset splitting and testing methodology. The abstract would also benefit from including crucial technical details such as the number of distinct face shape classes, specific transfer learning configurations (e.g., which layers were unfrozen, specific learning rates), and ideally, a comparison against other baseline models or state-of-the-art results for this particular face shape classification task. Addressing these points would significantly strengthen the paper's scientific rigor and overall impact.

Full Text

Comments

You need to be logged in to post a comment.