Towards interpretable & robust face recognition

• Main Author: Pattra, Surya Paryanta
• Other Authors: Lin Weisi
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision; Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/156565
• Institution: Nanyang Technological University

With the recent advancements of deep learning in computer vision, current state-of-the-art face recognition algorithms have surpassed human-level performance. However, they are not robust against constrained environments, especially image occlusions. To tackle the existing problem of occluded face recognition with facial masks, existing approaches utilize masks-detector module to detect and filter out the masks. In addition, those methods are trained using occluded version of datasets. Our proposed architecture, however, is able to be trained on general face dataset and generalize well into facial-masks occlusion. We also showed that our solution could surpass previous baselines.