MM-3DScene: 3D Scene Understanding by Customizing Masked Modeling with Informative-Preserved Reconstruction and Self-Distilled Consistency

This repository is built for the official implementation of:

MM-3DScene: 3D Scene Understanding by Customizing Masked Modeling with Informative-Preserved Reconstruction and Self-Distilled Consistency (CVPR2023) [arXiv],

by Mingye Xu*, Mutian Xu*, Tong He, Wanli Ouyang, Yali Wang†, Xiaoguang Han, Yu Qiao†.

Introduction

Masked Modeling (MM) has demonstrated widespread success in various vision challenges, by reconstructing masked visual patches. Yet, applying MM for large-scale 3D scenes remains an open problem due to the data sparsity and scene complexity. The conventional random masking paradigm used in 2D images often causes a high risk of ambiguity when recovering the masked region of 3D scenes.

To this end, we propose a novel informative-preserved reconstruction, which explores local statistics to discover and preserve the representative structured points, effectively enhancing the pretext masking task for 3D scene understanding. Integrated with a progressive reconstruction manner, our method can concentrate on modeling regional geometry and enjoy less ambiguity for masked reconstruction. Besides, such scenes with progressive masking ratios can also serve to self-distill their intrinsic spatial consistency, requiring to learn the consistent representations from unmasked areas. By elegantly combining informative-preserved reconstruction on masked areas and consistency self-distillation from unmasked areas, a unified framework called MM-3DScene is yielded.

We conduct comprehensive experiments on a host of downstream tasks. The consistent improvement (e.g., +6.1% [email protected] on object detection and +2.2% mIoU on semantic segmentation) demonstrates the superiority of our approach.

Usage

The codes and model zoos can be found in the folders of corresponding tasks.

For detection, follow the README under the detection folder.

For segmentation, follow the README under the segmentation folder.

Citing our work

If you use this code, please consider citing:

@inproceedings{xu2023mm3dscene,
  title={MM-3DScene: 3D Scene Understanding by Customizing Masked Modeling with Informative-Preserved Reconstruction and Self-Distilled Consistency},
  author={Xu, Mingye and Xu, Mutian and He, Tong and Ouyang, Wanli and Wang, Yali and Han, Xiaoguang and Qiao, Yu},
  booktitle={CVPR},
  year={2023}
}

Contact

You are welcome to send pull requests or share some ideas with us. Contact information: Mingye Xu ([email protected]) or Mutian Xu ([email protected]).

Acknowledgement

We include the following libraries and algorithms:
[1] CD
[2] PointTransformer
[3] VoteNet