|제목||[학술세미나] [학과세미나] 10월 12일(금) 11시 학과세미나 안내|
|내용||[학과세미나] 10월 12일(금) 11시 학과세미나 안내
▪제목 : Deep Learning based Image Processing: Supervised and Unsupervised Approaches
▪연사 : 전세영 (UNIST 전기전자컴퓨터공학부)
▪일시 : 2018년 10월 12일(금) AM 11:00 – 12:00
▪장소 : 25동 405호
Deep learning-based approaches have yielded state-of-the-art results in many high-level computer vision tasks such as image classification, object detection, scene understanding, and so on. Recently, many deep learning-based approaches have also been proposed and achieved outstanding performance in low-level computer vision tasks such as single image super resolution (SISR) and denoising. In this talk, I will briefly overview recent advances in supervised deep learning approaches for image processing such as SISR and denoising. Then, I will present EMBSR, our new image enhancement deep network that won the 3rd place in CVPR’18 NTIRE challenge in SISR Tracks 2,3, &4: Realistic, for images that were contaminated by multiple degradation sources. Secondly, I will review recent unsupervised approaches for image processing, especially focusing on training deep denoising networks WITHOUT clean ground truth images. Then, I will introduce our recent SURE (Stein’s Unbiased Risk Estimator) based approach that were able to achieve performance close to deep denoisers that were trained with ground truth data.
세미나 안내_181012_전세영.hwp [14.5KB]