Program


Dec. 3, 2020 (Beijing 8:00~11:10, 16:00~19:10 PST, 19:00~22:10 EST, 1:00~4:10 CET, 9:00~12:10 JST)

Session One (Beijing 8:00~9:30, 16:00~17:30 PST, 19:00~20:30 EST, 1:00~2:30 CET, 9:00~10:30 JST):

Beijing 8:00~8:30: Invited Talk (Dr. Takuhiro Kaneko)
Learning to Generate Images with Imperfect Supervision

Talk abstract: In computer vision and machine learning, image generation has been actively studied owing to its various applications, such as photo editing and image restoration. Recently, deep generative models, such as generative adversarial networks (GANs), have garnered attention owing to their high image reproduction ability. However, the limitations of the typical models are that they require complete supervision for training. For example, to create a model that can be controlled in detail, we need to collect the fine-grained annotation about the status. To construct a model that can synthesize clean images, we need to prepare a large-scale dataset that includes many clean photos. To overcome these limitations, we recently proposed several extensions of GANs that can learn to generate images with weak or noisy supervision. In this talk, I will introduce their core ideas and show their effectiveness with the experimental results.

Beijing 8:30~8:45
Wenhao Fang, Xian-hua Han, Spatial and Channel Attention Modulated Network for Medical Image Segmentation

Beijing 8:45~9:00
Oishee Hoque, Al-Farabi Akash, Md. Saiful Islam, Mohammad Imrul Jubair, BdSL36: A Dataset for Bangladeshi Sign Letters Recognition

Beijing 9:00~9:15
Zhe Liu, Yinqiang Zheng, Xian-Hua Han, Unsupervised Multispectral and Hyperspectral Image Fusion with Deep Spatial and Spectral Priors

Beijing 9:15~9:30
Kiran Akadas, Shankar Gangisetty, 3D Semantic Segmentation of Large-Scale Scene Understanding

Session two (Beijing 9:40~11:10, 17:40~19:10 PST, 20:40~22:10 EST, 2:40~4:10 CET, 10:40~12:10 JST):

Beijing 9:40~10:10: Invited Talk (Dr. Zheng Wang)
Is person re-ID a solved problem?

Talk abstract: Person re-identification (re-ID) is a CV technology that finds a certain person of interest in a large number of videos. It facilitates various applications that require painful and boring video watching, including searching for video shots related to an actor of interest from TV series, a lost child in a shopping mall from camera videos, a suspect from a city surveillance system. Its efficiency and effectiveness accelerate the process of video analysis. In recent years, existing technologies have only been evaluated on standard benchmarks. Although we have made significant advances in standard datasets, it is still far away to design an open-world re-identification system. In this talk, I will conduct a brief review of general person re-ID, where the person's appearance variation, the short-term environment change, and the intra-modality discrepancy work as the main challenge. I will introduce new trends of person re-ID system that more practical in open-world conditions, consisting of group, long-term, and cross-modality. Representative approaches, comparisons, and discussions will be given.

Beijing 10:10~10:25
Yinhao Li, Yutaro Iwamoto, Lanfen Lin, Yen-Wei Chen, Parallel-Connected Residual Channel Attention Network for Remote Sensing Image Super-Resolution

Beijing 10:25~10:40
Seiya Fujita, Xian-Hua Han, Cell Detection and Segmentation in Microscopy Images with Improved Mask R-CNN

Beijing 10:40~10:55
Yuan Zhuang, Lanfen Lin, Ruofeng Tong, Jiaqing Liu, Yutaro Iwamoto, Yen-Wei Che, G-GCSN: Global Graph Convolution Shrinkage Network for Emotion Perception from Gait

Beijing 10:55~11:10
Philipp Andermatt, Radu Timofte, A Weakly Supervised Convolutional Network For Change Segmentation and Classification