Program

CVM 2022 Conference Programme

	Thursday, April 7, 2022
08:30 - 18:00	Registration
	Jittor Course on Visual Media Learning
15:00 - 16:00	Jittor: Fundamentals and Lastest Progress Dun Liang (Tsinghua University)
16:00 - 17:00	Attention in Computer Vision Meng-Hao Guo (Tsinghua University)
17:00 - 18:00	Visual Media Editing and Generation with Neural Rendering Lin Gao (Institute of Computing Technology, Chinese Academy of Sciences)
Day One	Friday, April 8, 2022
08:00 - 08:30	Registration
08:30 - 09:00	Opening (Chair: Shi-Min Hu)
	Keynote talk I (Chair: Seungyong Lee)
09:00 - 09:40	Mind the gap: Learning 3D representation from 2D image collections Xin Tong (Microsoft Research Asia)
	Industrial Session (Chair: Hui Zhang)
09:40 - 10:00	Modern Game Engine Architecture Introduction Lei Su (CTO of Booming Technology)
10:00 - 10:20	Procedural Content Generation for Borderless Creation Haozhi Huang (Technical Partner of XVERSE Technology)
10:20 - 10:40	Coffee break
Session 1	Image Synthesis (Chair: Yong-Jin Liu)
10:40 - 10:55	STATE: Learning Structure and Texture Representations for Novel View Synthesis Xinyi Jing, Qiao Feng, Yu-Kun Lai, Jinsong Zhang, Yuanqiang Yu, Kun Li
10:55 - 11:10	A Comparative Study of CNN- and Transformer-based Neural Style Transfer Huapeng Wei, Yingying Deng, Fan Tang, Xingjia Pan, Weiming Dong
11:10 - 11:25	StrokeGAN Painter: Learning to Paint Artworks Using Stroke-Style Generative Adversarial Networks Qian Wang, Cai Guo, Hong-Ning Dai, Ping Li
11:25 - 11:40	Unsupervised Image Translation with Distributional Semantics Awareness Zhexi Peng, He Wang, Yanlin Weng, Yin Yang, Tianjia Shao
Session 2	Rendering (Chair: Kun Xu)
11:40 - 11:55	A Psychoacoustic Quality Criterion for Path-Traced Sound Propagation Chunxiao Cao, Zili An, Zhong Ren, Dinesh Manocha, Kun Zhou
11:55 - 12:10	Neural Temporal Denoising for Indirect Illumination Yan Zeng, Lu Wang, Yanning Xu, Xiangxu Meng
12:10 - 14:00	Lunch
Session 3	Geometry & Point Cloud (Chair: Yang Liu)
14:00 - 14:15	Out-of-core Outlier Removal for Large-scale Indoor Point Clouds Linlin Ge, Jieqing Feng
14:15 - 14:30	Towards Uniform Point Distribution in Feature-Preserving Point Cloud Filtering Shuanjun Chen, Jinxi Wang, Wei Pan, Shang Gao, Meili Wang, Xuequan Lu
14:30 - 14:45	Jacobi--PIA Algorithm for Bi-Cubic B-Spline Interpolation Surfaces Chengzhi Liu, Juncheng Li, Lijuan Hu
Session 4	Image Processing (Chair: Lei Zhang)
14:45 - 15:00	Light Field Super-Resolution Using Complementary-View Feature Attention Wei Zhang, Wei Ke, Da Yang, Hao Sheng, Zhang Xiong
15:00 - 15:15	Autocomplete Repetitive Stroking with Image Guidance Yilan Chen, Kin Chung Kwan, Hongbo Fu
15:15 - 15:30	Polygonal Finite Element Based Content-Aware Image Warping Juan Cao, Xiaoyi Zhang, Jiannan Huang, Yongjie Jessica Zhang
15:30 - 15:45	Coffee break
Session 5	Virtual Reality (Chair: Miao Wang)
15:45 - 16:00	ARSlice: Head-Mounted Display Augmented with Dynamic Tracking and Projection Yu-Ping Wang, Sen-Wei Xie, Lihui Wang, Hongjin Xu, Satoshi Tabata, Masatoshi Ishikawa
16:00 - 16:15	Adaptive Optimization Algorithm for Resetting Techniques in Obstacle-ridden Environments (invited TVCG paper) Song-Hai Zhang, Chia-Hao Chen, Zheng Fu, Yongliang Yang, Shi-Min Hu
Session 6	Shape Analysis (Chair: Lin Gao)
16:15 - 16:30	Learning-based Intrinsic Reflectional Symmetry Detection Yi-Ling Qiao, Gao Lin, Shu-Zhi Liu, Ligang Liu, Yu-Kun Lai, Xilin Chen
16:30 - 16:45	Deep Functional Maps for Simultaneously Computing Direct and Symmetric Correspondences of 3D Shapes Hui Wang, Bitao Ma, Junjie Cao, Xiuping Liu, Hui Huang
16:45 - 17:00	TAD-Net: tooth axis detection network based on rotation transformation encoding Yeying Fan, Qian Ma, Guangshun Wei, Zhiming Cui, Yuanfeng Zhou, Wenping Wang
Poster Session
17:00 - 18:00	Fuzzy-based Indoor Scene Modeling with Differentiated Examples Qiang Fu, Shuhan He, Zhigang Deng, Xueming Li, Hongbo Fu Deep Unfolding Multi-scale Regularizer Network for Image Denoising Jingzhao Xu, Mengke Yuan, Dongming Yan, Tieru Wu High-Quality Unsupervised Image Denoising via Multi-Scale Deep Image Prior Qing Zhang, Yongwei Nie, Lei Zhu, Wei-Shi Zheng Shape Embedding and Retrieval in Multi-Flow Deformation Baiqiang Leng, Jingwei Huang, Guanlin Shen, Bin Wang Deep Multi-Task Learning based Fingertip Detection Ruize Han, Jiewen Zhao, Liang Wan Shape-aware Stroke Segmentation for Calligraphic Characters Zibo Zhang, Xueting Liu, Chengze Li, Huisi Wu, Zhenkun Wen Attribute Consistency Guided Generative Adversarial Networks for Unsupervised Image-to-Image Translation Fengjiang Liu, Li Yao Point Cloud Completion on Structured Feature Map with Feedback Network Zejia Su, Haibin Huang, Chongyang Ma, Hui Huang, Ruizhen Hu Multi-foreground objects segmentation based on RGB-D Image Yan Li, Di Zhu, Hui Chen, Haikun Li, Changhe Tu Adaptive Content-aware Correction for Wide-angle Portrait Photos Juan Cao, Binyan Lin, Zhonggui Chen TransLoc3D: point cloud based large-scale place recognition using adaptive receptive fields Tian-Xing Xu, Yuan-Chen Guo, Yu-Kun Lai, Song-Hai Zhang Sphere Face Model:A 3D Morphable Model with Hypersphere Manifold Latent Space Diqiong Jiang, Yiwei Jin, Fang-Lue Zhang, Yun Zhang, Zhe Zhu, Ruofeng Tong, Min Tang
18:00 - 20:00	Conference Banquet
Day Two	Saturday, April 9, 2022
	Keynote talk II (Chair: Shi-Min Hu)
09:00 - 09:40	Data-efficient GAN Training Jun-Yan Zhu (Carnegie Mellon University)
Session 7	Attention (Chair: Weiming Dong)
09:40 - 09:55	Attention Mechanisms in Computer Vision: A Survey Meng-Hao Guo, Tian-Xing Xu, Jiang-Jiang Liu, Zheng-Ning Liu, Peng-Tao Jiang, Tai-Jiang Mu, Song-Hai Zhang, Ralph Martin, Ming-Ming Cheng, Shi-Min Hu
09:55 - 10:10	Self-supervised Coarse-to-fine Monocular Depth Estimation Using Lightweight Attention Module Yuanzhen Li, Fei Luo, Chunxia Xiao
10:10 - 10:25	Attention-based Dual Supervised Decoder for RGBD Semantic Segmentation Yang Zhang, Yang Yang, Chenyun Xiong, Guodong Sun, Yanwen Guo
10:25 - 10:40	Coffee break
Session 8	Meshes & 3D Printing (Chair: Lin Gao)
10:40 - 10:55	Patch-based mesh inpainting via low rank recovery Xiaoqun Wu, Xiaoyun Lin, Nan Li, Haisheng Li
10:55 - 11:10	Untangling All-Hex Meshes via Adaptive Boundary Optimization Qing Huang, Wen-Xiang Zhang, Qi Wang, Ligang Liu, Xiao-Ming Fu
11:10 - 11:25	3D Printed Hair Modeling from Strand-level Hairstyles Han Chen, Minghai Chen, Lin Lu
Session 9	Understanding (Chair: Shi-Sheng Huang)
11:25 - 11:40	Element-Arrangement Context Network for Facade Parsing Yan Tao, Yiteng Zhang, Xuejin Chen
11:40 - 11:55	Probability-based channel pruning for depthwise separable convolutional networks Hanli Zhao, Kaijie Shi, Xiaogang Jin, Mingliang Xu, Hui Huang, Wanglong Lu, Ying Liu
11:55 - 12:10	Learn Robust Pedestrian Representation within Minimal Modality Discrepancy for Visible-Infrared Person Re-Identification Yujie Liu, Wenbin Shao, Xiaorui Sun
12:10 - 14:00	Lunch
	Keynote talk III (Chair: Hongbo Fu)
14:00 - 14:40	Human-in-the-Loop Preferential Bayesian Optimization for Visual Design Yuki Koyama (AIST)
Session 10	Face (Chair: Tianjia Shao)
14:40 - 14:55	3D-CariGAN: An End-to-End Solution to 3D Caricature Generation from Normal Face Photos (invited TVCG paper) Zipeng Ye, Mengfei Xia, Yanan Sun, Ran Yi, Minjing Yu, Juyong Zhang, Yu-Kun Lai, Yong-Jin Liu
14:55 - 15:10	Towards Harmonized Regional Style Transfer and Manipulation for Facial Images Cong Wang, Fan Tang, Yong Zhang, Weiming Dong, Tieru Wu
15:10 - 15:25	Learning Physically-based Material and Lighting Decompositionsfor Face Editing Qian Zhang, Vikas Thamizharasan, James Tompkin
15:25 - 15:40	Coffee break
Session 11	Simulation & Visualization (Chair: Bo Ren)
15:40 - 15:55	Simulating Fractures with Bonded Discrete Element Method (invited TVCG paper) Jia-Ming Lu, Chenfeng Li, Geng-Chen Cao, Shi-Min Hu
15:55 - 16:10	O3NJ Trees: Optimally Ordered Orthogonal Neighbor Joining Trees for Hierarchical Cluster Analysis Tong Ge, Yunhai Wang, Michael Sedlmair, Zhanglin Cheng, Ying Zhao, Xin Liu, Baoquan Chen, Oliver Deussen
Session 12	Tracking & SLAM (Chair: Paul Rosin)
16:10 - 16:25	Local Homography Estimation on User-specified Textureless Regions Zheng Chen, Xiaonan Fang, Songhai Zhang
16:25 - 16:40	CGTracker: Center Graph Network for One-Stage Multi-Object Detection and Tracking Xin Feng, Haoming Wu, Yihao Yin, Yongbo Li, Libin Lan
16:40 - 16:55	ObjectFusion: Accurate Object-level SLAM with Neural Object Priors Zi-Xin Zou, Shi-Sheng Huang, Tai-Jiang Mu, Yu-Ping Wang
16:55 - 17:25	Closing Session

Keynote Speakers

Xin Tong, Microsoft Research Asia

Title:

Mind the gap: Learning 3D representation from 2D image collections

Abstract:

3D deep learning has demonstrated its advantage in many 3D graphics applications. However, compared to images and videos that can be easily acquired from real world, modeling or capturing 3D dataset (e.g. shapes and material maps) is still a difficult task, which limits the scale of 3D dataset available in 3D deep learning.
In this talk, I will introduce our explorations in the last several years on how to utilize 2D image collections in 3D deep learning. By bridging the gap between 2D images and 3D representations, we believe that this method will release the power of deep learning and enable new solutions for 3D content creation.

Speaker's Biography:

Dr. Xin Tong now is a partner research manager of Microsoft Research Asia (MSRA) and the leader of graphics group in MSRA. His research interests cover many topics in computer graphics and computer vision, including appearance modeling and rendering, texture synthesis, light transport analysis, 3D deep learning, performance capturing and facial animation, as well as graphics system. Xin has published more than 150 papers in top computer graphics and vision journals and conferences, including 55 SIGGRAPH/TOG papers. He has served as the associate editor of computer graphics journals (ACM TOG, IEEE TVCG, CGF) and paper committee members of ACM SIGGRAPH/SIGGRAPH ASIA, Eurographics, and Pacific Graphics. He is the associate editor of IEEE CG&A, CVMJ, and visual informatics. Xin obtained his Ph.D. degree in Computer Graphics from Tsinghua University in 1999 and his B.S. Degree and Master Degree in Computer Science from Zhejiang University in 1993 and 1996 respectively.

---

Jun-Yan Zhu, Carnegie Mellon University

Title:

Data-efficient GAN Training

Abstract:

The power and promise of deep generative models such as GANs lie in their ability to synthesize endless realistic, diverse, and novel visual content. Unfortunately, the creation and deployment of these large-scale GANs demand high-performance computing platforms and large-scale annotated datasets. Commonly used datasets such as ImageNet and LSUN require human annotation of millions of images. In this talk, I will present two data-efficient GAN training methods: differentiable data augmentation and ensembling off-the-shelf computer vision models. Collectively, these techniques allow us to learn a high-quality GAN model with as few as one hundred photos. If time permits, I will also discuss the issues of existing GANs evaluation metrics as well as potential fixes.

Speaker's Biography:

Jun-Yan Zhu is an Assistant Professor at the School of Computer Science of Carnegie Mellon University. Prior to joining CMU, he was a Research Scientist at Adobe Research and a postdoctoral researcher at MIT CSAIL. He obtained his Ph.D. from UC Berkeley and his B.E. from Tsinghua University. He studies computer vision, computer graphics, computational photography, and machine learning. He is the recipient of the Facebook Fellowship, ACM SIGGRAPH Outstanding Doctoral Dissertation Award, and UC Berkeley EECS David J. Sakrison Memorial Prize for outstanding doctoral research. His co-authored work has received the NVIDIA Pioneer Research Award, SIGGRAPH 2019 Real-time Live Best of Show Award and Audience Choice Award, and The 100 Greatest Innovations of 2019 by Popular Science.

---

Yuki Koyama, National Institute of Advanced Industrial Science and Technology (AIST)

Title:

Human-in-the-Loop Preferential Bayesian Optimization for Visual Design

Abstract:

Visual design often involves searching for the best parameter set that yields the best-looking design. However, such tasks are difficult to solve with typical optimization algorithms since the objective function is based on subjective evaluation and thus requires special treatment. This talk will introduce preferential Bayesian optimization (PBO), a powerful technique for handling such subjective tasks. PBO is a human-in-the-loop Bayesian optimization method that runs with relative preferential evaluation (e.g., which design is the best compared among several alternatives) instead of absolute evaluation (e.g., how good the current design is). This technique constructs a predictive model of the latent preference and generates effective preference queries to human evaluators based on the predictive model. Then, I will explain two advanced PBO methods [SIGGRAPH 2017; SIGGRAPH 2020] that achieve even better sample efficiency by combining with tailored user interactions.

Speaker's Biography:

Dr. Yuki Koyama is a Researcher at the National Institute of Advanced Industrial Science and Technology (AIST). He received his Ph.D. from the University of Tokyo in 2017, advised by Prof. Takeo Igarashi. His research fields are computer graphics and human-computer interaction, and he has published his first-authored papers at top venues such as SIGGRAPH, SIGGRAPH Asia, CHI, and UIST. His interest includes computational design and human-in-the-loop design optimization. From 2021, he also started working at Graphinica (a Japanese animation studio), in which he is aiming at bridging art and technology in animation production. He was awarded JSPS Ikushi Prize (2017) and Asiagraphics Young Researcher Award (2021).

---