Rening Image Representation With Click-Through Data", to appear in going to submmit in arXiv. Research Experience Research Assistant, Seppresent Communication and Multimedia Lab (CMLab), NTU Proposed a Convolutional Neural Network (CNN) accurately performing fine-grained classification for surveillance car. The model recovered lost details from low resolution image with hints in crawled web images. Leveraged frame similarity to speed up CNN-based object detection and semantic segmentation models, e.g. FPN, PSPNet, YOLO, Faster-RCNN, and SegNet. Proposed a multimodal CNN achieving high fine-grained classification accuracy. The model was trained with
國立台灣大學・
Electrical Engineering