Hybrid 3D Representations

My photo

Machine learning algorithms operate on vectorized data. When applying machine learning algorithms to various applications that involve 3D geometry, it is critical to developing suitable 3D representations. This problem is fundamental as there are many possible shape representations, e.g., the triangle mesh, point cloud, implicit surface, parametric surface, part-based models, and light-field representations. Instead of focusing on a single representation, I study hybrid 3D representations that combine the strength of different 3D representations. Key advantages include powerful feature extraction and unsupervised learning of neural networks.

Our early results have revealed the huge potential of hybrid 3D representations. [CVPR19] and [TOG20] show that utilizing hybrid representations can boost the performance of 3D semantic segmentation and 3D scene synthesis, respectively. [CVPR20a] and [CVPR20b] utilized hybrid representations for pose regression and relative pose regression, respectively. In particular, [CVPR20a] improved the ADD(~S) score on Occlusion-Limond by more than 60%.

In [ECCV2020], we introduced H3DNet, an approach for detecing 3D objects in 3D scenes. The basic idea is to fit 3D bounding boxes to an over-complete and hybrid set geometric primitives, including box center, box face centers, and box edge centers. The approach achieved state-of-the-art results on ScanNet and SUNRGBD benchmark datasets. In [ICCV21c], we introduced a primitive shape detection approach that combines semantic features, geometric features, and edge features, for segmenting a point cloud of a CAD model into primitive shapes. In [ICCV21b], we introduced a new scene synthesis approach that combines the synthesis of object attributes and object pair attributes. In [ICCV21a], we introduce a new approach for learning deformable shape generators by combing a data term and a novel deformation regularization loss. Our [ARXIV22] paper combines an implicit decoder and a point cloud encoder to learn point-based features for self-supervised learning of point clouds.

On a theoretic side, we study principled self-supervision constraints for unsupervised learning. In [CVPR19], we introduced path-invariant bases for joint learning a collection of neural networks that forms a directed graph. In [NeurIPS19], we introduced cycle-consistency bases among an undirected graph and an algorithm for optimizing cycle-consistency bases via a condition number of the training loss.

GenCorres_2023

[ARIV23] Haitao Yang, Xiangru Huang, Bo Sun, Chandrajit Bajaj and Qixing Huang. 3D Feature Prediction for Masked-AutoEncoder-Based Point Cloud Pretraining. arXiv preprint arXiv:2304.10523

IAE_2022

[ICCV23] Siming Yan, Zhenpei Yang, Haoxiang Li, Li Guan, Hao Kang, Gang Hua,and Qixing Huang. Implicit Autoencoder for Point Cloud Self-supervised Representation Learning. International Conference on Computer Vision (ICCV) 2023

ARAP_2021

[ICCV21a] Bo Sun, Xiangru Huang, , Zaiwei Zhang, Junfeng Jiang, Qixing Huang, and Chandrajit Bajaj. ARAPReg: An As-Rigid-As Possible Regularization Loss for Learning Deformable Shape Generators. International COnference on Computer Vision (or ICCV) 2021.

Synthesis_2021

[ICCV21b] Haitao Yang, Zaiwei Zhang, Siming Yan, Chongyang Ma, Haibin Huang, Yi Zheng, Chandrajit Bajaj, and Qixing Huang. Scene Synthesis via Uncertainty-Driven Attribute Synchronization. International Conference on Computer Vision (or ICCV) 2021.

HPNet_2021

[ICCV21c] Siming Yan, Zhenpei Yang, Chongyang Ma, Haibin Huang, Etienne Vouga, and Qixing Huang. HPNet: Deep Primitive Segmentation Using Hybrid Representations. International Conference on Computer Vision (or ICCV) 2021.

Arxiv_H3DNet

[ECCV20] Zaiwei Zhang, Bo Sun, Haitao Yang, and Qixing Huang. H3DNet: 3D Object Detection Using Hybrid Geometric Primitives. European Conference on COMPUTER VISION ( ECCV ) 2020.

tog_scene

[TOG20] Zaiwei Zhang, Zhenpei Yang, Chongyang Ma, Linjie Luo, Alexander Huth, Etienne Vouga, and Qixing Huang. Deep Generative Modeling for Scene Synthesis via Hybrid Representations ACM Transactions on Graphics, 39(2), 2020.

hybrid_pose

[CVPR20a] Chen Song, Jiaru Song, and Qixing Huang. HybridPose: 6D Object Pose Estimation under Hybrid Representations Computer Vision and Pattern Recognition (or CVPR) 2020.

hybrid_rel_pose

[CVPR20b] Zhenpei Yang, Siming Yan, and Qixing Huang. Extreme Relative Pose Network under Hybrid Representations Computer Vision and Pattern Recognition (or CVPR) 2020. Oral Presentation .

cycle_cons

[NeurIPS19] Leonidas Guibas, Qixing Huang. and Zhenxiao Liang. A Condition Number for Joint Optimization of Cycle-Consistent Networks. Advances in Neural Information Processing Systems(NIPS), 2019. Spotlight Presentation

arxiv_path

[CVPR19] Zaiwei Zhang, Zhenxiao Liang, Lemeng Wu, Xiaowei Zhou, and Qixing Huang. Path-Invariant Map Networks. Computer Vision and Pattern Recognition (or CVPR) 2019. Oral Presentation and Best Paper Finalist .