Program
Dec. 4, 2022 (Beijing 9:00~12:40, AM GMT+8)
Session One (Beijing 9:00~10:45):
9:00~9:30: Invited Talk (Dr. Boxin Shi, Peking University, China)
NeurImg: Hybrid Imaging Fusing Neuromorphic and Conventional Cameras
Talk abstract: In recent years, some specially designed neuromorphic cameras, such as event cameras and spike cameras, have drawn increasing attention of researchers for visual semantic analysis. Neuromorphic cameras have unique features different from conventional frame-based cameras, and they are particularly good at sensing very fast motion and high dynamic range scenes. In this talk, I will introduce our hybrid imaging framework fusing neuromorphic and conventional cameras. Such a system complements speed and dynamic range advantages of neuromorphic cameras with advantages in resolution, robustness, and color from RGB cameras, to achieve joint filtering of intensity images and neuromorphic events/spikes for high-resolution noise-robust imaging and high frame rate imaging without rolling shutter effect.
9:30~9:45
Zhe Liu; Xian-hua Han, Deep RGB-driven Learning Network for Unsupervised Hyperspectral Image Super-resolution
9:45~10:00
Daniel Steininger; Andreas Kriegler; Wolfgang Pointner; Verena Widhalm; Julia Simon; Oliver Zendel, Towards Scene Understanding for Autonomous Operations on Airport Aprons
10:00~10:15
Gan Zhan; Fang Wang; Weibin Wang; Yinhao Li; Qingqing Chen; Hongjiu Hu; Yen-Wei Chen, A Transformer-based Model for Preoperative Early Recurrence Prediction of Hepatocellular Carcinoma with Muti-phase MRI
10:15~10:30
Jiaqi Ning; Fei Li; Rujie Liu; Shun Takeuchi; Genta Mr Suzuk, Temporal Extension Topology Learning for Video-based Person Re-Identification
10:30~10:45
Zijia Wang; Wenbin Yang; Zhi-Song Liu; jiacheng ni; Qiang Chen; Zhen Jia, Gift from nature: Potential Energy Minimization for explainable dataset distillation
Session two (Beijing 11:00~12:30):
11:00~11:30: Invited Talk (Dr. Hong Liu, NII, Japan)
Evaluation of Person Re-identification Robustness: Attack and Defense
Talk abstract: Deep models appear to be susceptible to small, imperceptible changes over test instances, despite their remarkable success in person re-identification (ReID) applications. In this talk, I will provide a brief overview of recent advances in robust person ReID, such as attacks, defenses, and beyond. First, I will introduce a typical universal adversarial attack for general image retrieval. Then, I will discuss some recent findings about how meta-learning and virtual data can affect the person ReID robustness, including a holistic attack-defense framework along with virtual-guided meta-learning. Finally, I will present our work on the robust model under the cloth-changing scenario. I will conclude my talk with some insights into robust person ReID.
11:30~11:45
Kazuhiro Yamawaki; Xian-Hua Han, Lightweight Hyperspectral Image Reconstruction Network with Deep Feature Hallucination
11:45~12:00
Patrick Rim; Snigdha Saha; Marcus Rim, CaltechFN: Distorted and Partially Occluded Digits
12:00~12:15
Yongqing Sun; Xiaomeng Wu; Yukihiro Bandoh; Masaki Kitahara, Aerial Image Segmentation via Noise Dispelling and Content Distilling
12:15~12:30
Zijia Wang; Wenbin Yang; Zhi-Song Liu; jiacheng ni; Qiang Chen; Zhen Jia, OBJECT-CENTRIC Point Sets Feature Learning with Matrix Decomposition
12:30~12:40: Award and Closing Remarks