Multi-Channel Weight Sharing Convolutional Neural Network for Super-Resolution and Face Verification

• Main Authors: LAI, YI-CHEN, 賴怡辰
• Other Authors: CHIANG, CHEN-KUO
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/8v3zs2

碩士 === 國立中正大學 === 資訊工程研究所 === 105 === In computer vision, we usually need to find out the similarity and difference between images on verification and recognition problems. How to learn and extract the common features of the same identity is important. Inspired by Siamese Neural Network, we observe and design a deep learning architecture that utilizes weight sharing concept in a Convolutional Neural Network (CNN). We verify that it is meaningful and efficient to design shared weight deep learning architecture in various problems. For example, we apply weight sharing layers to multiple-channel super-resolution architecture for multiple low-resolution images and on our dual-channel architecture for face verification problem. The experimental results demonstrate the effectiveness of the proposed method on both the image enhancement in super-resolution and the accuracy of face verification.