Equivariant Vision: From Theory to Practice

CVPR 2024 Workshop, Seattle, WA
June 18 2024, 08:30-17:30

Exploiting symmetry in structured data is a powerful way to improve the generalization ability, data efficiency, and robustness of AI systems, which leads to the research direction of equivariant deep learning. Showing its effectiveness, it has been widely adopted in a large variety of subareas of computer vision, from 2D image analysis to 3D perception, as well as further applications such as medical imaging and robotics.

Our topics include but are not limited to:

  • Theoretical foundations of equivariant deep learning with symmetry and group theory.
  • Equivariance by design: Neural network architectures and mathematical guarantees.
  • Equivariance from data: Learning equivariant and invariant features.
  • Applications in 2D and 3D computer vision and robotics.
  • Applications in broader science: computational biology, medicine, natural science, etc.
  • Equivariance in the large-model era and potential future directions.

Keynote Speakers
Leonidas Guibas
Stanford
Haggai Maron
Technion & NVIDIA
Carlos Esteves
Google
Erik Bekkers
UvA
Nina Miolane
UCSB
Tutorial
Stefanos Pertigkiozoglou
UPenn
Evangelos Chatzipantazis
UPenn
Schedule
Session 1: Equivariance Theory and Network Design 08:30-11:45
Opening Remarks and Welcome 08:30-08:45
Keynote Talk: Haggai Maron
Exploiting Symmetries for Learning in Deep Weight Spaces
08:45-09:30
Keynote Talk: Leonidas Guibas
Learning and Enforcing Equivariance
09:30-10:15
Coffee Break 10:15-10:30
Keynote Talk: Erik Bekkers
Neural Ideograms and Geometry-Grounded Representation Learning
10:30-11:15
Spotlight Talk
LeaF: Learning Frames for 4D Point Cloud Sequence Understanding
11:15-11:30
Spotlight Talk
EquiAdapt: Equivariant Adaptation of Large Pretrained Models
11:30-11:45
Lunch Break 11:45-12:30
Accepted Paper Poster Session 12:30-14:00
Session 2: Applications in Computer Vision and Beyond 14:00-17:30
Keynote Talk: Carlos Esteves
Geometric Deep Learning for Weather
14:00-14:45
Keynote Talk: Nina Miolane
Hierarchical G-Equivariance in Vision
14:45-15:30
Coffee Break 15:30-15:45
Spotlight Talk
Diffusion-EDFs: Bi-equivariant Denoising Generative Modeling on SE(3) for Visual Robotic Manipulation
15:45-16:00
Spotlight Talk
DiffAssemble: A Unified Graph-Diffusion Model for 2D and 3D Reassembly
16:00-16:15
Tutorial: Getting Started with Equivariant Networks 16:15-17:15
Conclusion 17:15-17:30
Accepted Papers
(Spotlight) LeaF: Learning Frames for 4D Point Cloud Sequence Understanding
Yunze Liu, Junyu Chen, Zekai Zhang, Jingwei Huang, Li Yi
SERNet-Former: Semantic Segmentation by Efficient Residual Network with Attention-Boosting Gates and Attention-Fusion Networks
Serdar Erisen
[Poster]
Living Scenes: Multi-object Relocalization and Reconstruction in Changing 3D Environments
Liyuan Zhu, Shengyu Huang, Konrad Schindler, Iro Armeni
Making Vision Transformers Truly Shift-Equivariant
Renan A. Rojas-Gomez, Teck-Yian Lim, Minh N. Do, Raymond A. Yeh
[Poster]
Stability Analysis of Equivariant Convolutional Representations Through The Lens of Equivariant Multi-layered CKNs
Soutrik Roy Chowdhury
TetraSphere: A Neural Descriptor for O(3)-Invariant Point Cloud Analysis
Pavlo Melnyk, Andreas Robinson, Michael Felsberg, Mårten Wadenbäck
[Poster]
SurfelReloc: Surfel-based 3D Registration with Equivariant Features
Xueyang Kang, Hang Zhao, Zhaoliang Luan, Patrick Vandewalle, Kourosh Khoshelham
[Poster]
Improving Equivariance in State-of-the-Art Supervised Depth and Normal Predictors
Yuanyi Zhong, Anand Bhattad, Yu-Xiong Wang, David Forsyth
(Spotlight) Diffusion-EDFs: Bi-equivariant Denoising Generative Modeling on SE(3) for Visual Robotic Manipulation
Hyunwoo Ryu, Jiwoo Kim, Hyunseok An, Junwoo Chang, Joohwan Seo, Taehan Kim, Yubin Kim, Chaewon Hwang, Jongeun Choi, Roberto Horowitz
[Poster]
Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform
Chunghyun Park, Seungwook Kim, Jaesik Park, Minsu Cho
[Poster]
Neural Processing of Tri-Plane Hybrid Neural Fields
Adriano Cardace, Pierluigi Zama Ramirez, Francesco Ballerini, Allan Zhou, Samuele Salti, Luigi di Stefano
[Poster]
(Spotlight) EquiAdapt: Equivariant Adaptation of Large Pretrained Models
Arnab Kumar Mondal, Siba Smarak Panigrahi, Sékou-Oumar Kaba, Sai Rajeswar, Siamak Ravanbakhsh
[Poster]
RiEMann: Near Real-Time SE(3)-Equivariant Robot Manipulation without Point Cloud Segmentation
Chongkai Gao, Zhengrong Xue, Shuying Deng, Tianhai Liang, Siqi Yang, Lin Shao, Huazhe Xu
[Poster]
Equivariance versus Augmentation for Spherical Images
Jan E Gerken, Oscar Carlsson, Hampus Linander, Fredrik Ohlsson, Christoffer Petersson, Daniel Persson
[Poster]
Leveraging SE(3) Equivariance for Learning 3D Geometric Shape Assembly
Ruihai Wu, Chenrui Tie, Yushi Du, Yan Shen, Hao Dong
[Poster]
EqvAfford: SE(3) Equivariance for Point-Level Affordance Learning
Yue Chen, Chenrui Tie, Ruihai Wu, Hao Dong
[Poster]
Color Equivariant Network
Felix O'Mahony, Yulong Yang, Christine Allen-Blanchette
[Poster]
(Spotlight) DiffAssemble: A Unified Graph-Diffusion Model for 2D and 3D Reassembly
Stefano Fiorini, Gianluca Scarpellini, Francesco Giuliari, Pietro Morerio, Alessio Del Bue
[Poster]
Steerers: A framework for rotation equivariant keypoint descriptors
Georg Bökman, Johan Edstedt, Michael Felsberg, Fredrik Kahl
[Poster]
Improved Canonicalization for Model Agnostic Equivariance
Siba Smarak Panigrahi, Arnab Kumar Mondal
Organizers
Congyue Deng
Stanford
Jiahui Lei
UPenn
Yinshuang Xu
UPenn
Li Yi
Tsinghua
Christine Allen-Blanchette
Princeton
Vitor Guizilini
TRI
Ameesh Makadia
Google
Kostas Daniilidis
UPenn
Top