Table of contents

1. Installation guide
2. Data Preparation
3. Run the code
4. Acknowledgments

Unofficial Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking CVPR'25

Abstract: We introduce Any3DIS, a novel class-agnostic approach for 3D instance segmentation that leverages 2D mask tracking to segment 3D objects in point cloud scenes. Existing 3D instance segmentation methods frequently encounter issues with over-segmentation, leading to redundant and inaccurate 3D proposals that complicate downstream tasks. This challenge arises from their unsupervised merging approach, where dense 2D instance masks are lifted across frames into point clouds to form 3D candidate proposals without direct supervision. These candidates are then hierarchically merged based on heuristic criteria, often resulting in numerous redundant segments that fail to combine into precise 3D proposals. To overcome these limitations, we propose a 3D-Aware 2D Mask Tracking module that uses robust 3D priors from a 2D mask segmentation and tracking foundation model (SAM-2) to ensure consistent object masks across video frames. Rather than merging all visible superpoints across views to create a 3D mask, our 3D Mask Optimization module leverages a dynamic programming algorithm to select an optimal set of views, refining the superpoints to produce a final 3D proposal for each object. Our approach achieves comprehensive object coverage within the scene while reducing unnecessary proposals, which could otherwise impair downstream applications. Evaluations on ScanNet200 and ScanNet++ confirm the effectiveness of our method, with improvements across Class-Agnostic, Open-Vocabulary, and Open-Ended 3D Instance Segmentation tasks.

@inproceedings{nguyen2025any3dis,
  title={Any3DIS: Class-Agnostic 3D Instance Segmentation by 2D Mask Tracking},
  author={Nguyen, Phuc and Luu, Minh and Tran, Anh and Pham, Cuong and Nguyen, Khoi},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={3636--3645},
  year={2025}
}

Features 📣

• State-of-the-art performance of Class-Agnostic 3D Instance Segmentation on ScanNet200 and ScanNet++.
• State-of-the-art performance of Open-Vocabulary 3D Instance Segmentation on ScanNet200 and ScanNet++.
• State-of-the-art performance of Open-Ended 3D Instance Segmentation on ScanNet200 and ScanNet++.
• Support Interactive 3D Instance Segmentation.
• Reproducibility code for ScanNet200, Scannet++ datasets!
• 2D segmenter: Support SAM, SAM-2!

Installation guide 🔨

Please refer to installation guide

Data Preparation 📂

Please refer to data preparation guide

Run the code 🏃

Please refer to running the code guide

Acknowledgments

This repo is built upon Open3DIS.

Contacts

If you have any questions or suggestions about this repo, please feel free to contact me (phucnda@gmail.com).