mlsanityData to 7 Oct 2026

UnScene3D: Unsupervised 3D Instance Segmentation for Indoor Scenes

arXiv 2303.14541 ↗First posted 25 Mar 2023cs.CVcited by 31

Not on the map. Why?

Abstract (arXiv)

3D instance segmentation is fundamental to geometric understanding of the world around us. Existing methods for instance segmentation of 3D scenes rely on supervision from expensive, manual 3D annotations. We propose UnScene3D, the first fully unsupervised 3D learning approach for class-agnostic 3D instance segmentation of indoor scans. UnScene3D first generates pseudo masks by leveraging self-supervised color and geometry features to find potential object regions. We operate on a basis of geometric oversegmentation, enabling efficient representation and learning on high-resolution 3D data. The coarse proposals are then refined through self-training our model on its predictions. Our approach improves over state-of-the-art unsupervised 3D instance segmentation methods by more than 300% Average Precision score, demonstrating effective instance segmentation even in challenging, cluttered 3D scenes.

Key facts. Cited by 31 papers in the map (data through 7 Oct 2026). Builds on 42 of its 70 references that are in the map. First posted on arXiv 25 Mar 2023.

Builds on

42 of 70 references.

References · 42

  1. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  2. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,384
  3. 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,088
  4. 2018Dynamic Graph CNN for Learning on Point Cloudscited by 2,220
  5. 20194D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networkscited by 1,201
  6. 20173D Semantic Segmentation with Submanifold Sparse Convolutional Networkscited by 746
  7. 2018PointConv: Deep Convolutional Networks on 3D Point Cloudscited by 656
  8. 2017Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentationscited by 614
  9. 2019RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Cloudscited by 584
  10. 2022OpenScene: 3D Scene Understanding with Open Vocabulariescited by 556
  11. 2020PointContrast: Unsupervised Pre-training for 3D Point Cloud Understandingcited by 502
  12. 2021ARKitScenes: A Diverse Real-World Dataset For 3D Indoor Scene Understanding Using Mobile RGB-D Datacited by 442
30 more
  1. 2023ConceptFusion: Open-set Multimodal 3D Mappingcited by 302
  2. 2022Language-Grounded Indoor 3D Semantic Segmentation in the Wildcited by 243
  3. 2021Self-Supervised Pretraining of 3D Features on any Point-Cloudcited by 217
  4. 20183D-SIS: 3D Semantic Instance Segmentation of RGB-D Scanscited by 216
  5. 2020Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contextscited by 210
  6. 2022SoftGroup for 3D Instance Segmentation on Point Cloudscited by 197
  7. 2023Cut and Learn for Unsupervised Object Detection and Instance Segmentationcited by 186
  8. 2020OccuSeg: Occupancy-aware 3D Instance Segmentationcited by 177
  9. 2022PLA: Language-Driven Open-Vocabulary 3D Scene Understandingcited by 176
  10. 2017SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentationcited by 168
  11. 2022Self-Supervised Transformers for Unsupervised Object Discovery using Normalized Cutcited by 166
  12. 2022CLIP-Fields: Weakly Supervised Semantic Fields for Robotic Memorycited by 164
  13. 2019Learning Object Bounding Boxes for 3D Instance Segmentation on Point Cloudscited by 156
  14. 2018GSPN: Generative Shape Proposal Network for 3D Instance Segmentation in Point Cloudcited by 145
  15. 2021Hierarchical Aggregation for 3D Instance Segmentationcited by 126
  16. 20183DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentationcited by 118
  17. 2022PartSLIP: Low-Shot Part Segmentation for 3D Point Clouds via Pretrained Image-Language Modelscited by 117
  18. 2022Superpoint Transformer for 3D Scene Instance Segmentationcited by 115
  19. 2022FreeSOLO: Learning to Segment Objects without Annotationscited by 102
  20. 2017Accelerated Hierarchical Density Clusteringcited by 100
  21. 2021Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networkscited by 98
  22. 2021One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentationcited by 89
  23. 2019Multi-view PointNet for 3D Scene Understandingcited by 76
  24. 2018Fully-Convolutional Point Networks for Large-Scale Point Cloudscited by 44
  25. 2023Top-Down Beats Bottom-Up in 3D Instance Segmentationcited by 31
  26. 2019RevealNet: Seeing Behind Objects in RGB-D Scanscited by 30
  27. 2022OGC: Unsupervised 3D Object Segmentation from Rigid Dynamics of Point Cloudscited by 24
  28. 2023Mask3D: Pre-training 2D Vision Transformers by Learning Masked 3D Priorscited by 15
  29. 2019Incremental Class Discovery for Semantic Segmentation with RGBD Sensingcited by 13
  30. 2022One Class One Click: Quasi Scene-level Weakly Supervised Point Cloud Semantic Segmentation with Active Learningcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 7; 2024: 7; 2025: 12; 2026: 5.

31 papers.

Sort:

By area:

  1. 2023SAI3D: Segment Any Instance in 3D Scenescited by 90
  2. 2023OpenIns3D: Snap and Lookup for 3D Open-vocabulary Instance Segmentationcited by 82
  3. 2023SAMPro3D: Locating SAM Prompts in 3D for Zero-Shot Instance Segmentationcited by 44
  4. 2024EmbodiedSAM: Online Segment Any 3D Thing in Real Timecited by 40
  5. 2023Lowis3D: Language-Driven Open-World Instance-Level 3D Scene Understandingcited by 38
  6. 2023SAM-guided Graph Cut for 3D Instance Segmentationcited by 35
  7. 2025Spatial Understanding from Videos: Structured Prompts Meet Simulation Datacited by 28
  8. 2024Part2Object: Hierarchical Unsupervised 3D Instance Segmentationcited by 12
  9. 2024Articulate3D: Holistic Understanding of 3D Scenes as Universal Scene Descriptioncited by 9
  10. 2025Inpaint360GS: Efficient Object-Aware 3D Inpainting via Gaussian Splatting for 360deg Scenescited by 8
  11. 2023FreePoint: Unsupervised Point Cloud Instance Segmentationcited by 6
  12. 2025Pointmap Association and Piecewise-Plane Constraint for Consistent and Compact 3D Gaussian Segmentation Fieldcited by 6
19 more
  1. 2024UNIT: Unsupervised Online Instance Segmentation through Timecited by 5
  2. 2025Masked Scene Modeling: Narrowing the Gap Between Supervised and Self-Supervised Learning in 3D Scene Understandingcited by 4
  3. 2025Sketchy Bounding-box Supervision for 3D Instance Segmentationcited by 4
  4. 2024AutoInst: Automatic Instance-Based Segmentation of LiDAR 3D Scanscited by 3
  5. 2024DCSEG: Decoupled 3D Open-Set Segmentation using Gaussian Splattingcited by 3
  6. 2025GrabS: Generative Embodied Agent for 3D Object Segmentation without Scene Supervisioncited by 3
  7. 2025MoonSeg3R: Monocular Online Zero-Shot Segment Anything in 3D with Reconstructive Foundation Priorscited by 3
  8. 2025Easy3D: A Simple Yet Effective Method for 3D Interactive Segmentationcited by 2
  9. 2025SGS-3D: High-Fidelity 3D Instance Segmentation via Reliable Semantic Mask Splitting and Growingcited by 2
  10. 2023Generalized Robot 3D Vision-Language Model with Fast Rendering and Pre-Training Vision-Language Alignmentcited by 1
  11. 2024Semantic UV mapping to improve texture inpainting for indoor scenescited by 1
  12. 2025Hierarchical Image-Guided 3D Point Cloud Segmentation in Industrial Scenes via Multi-View Bayesian Fusioncited by 1
  13. 2026SpaceEra++: A Unified Framework Towards 3D Spatial Reasoning in Videocited by 1
  14. 2025Unsupervised Online 3D Instance Segmentation with Synthetic Sequences and Dynamic Losscited by 0
  15. 2025Class-agnostic 3D Segmentation by Granularity-Consistent Automatic 2D Mask Trackingcited by 0
  16. 2026FoundObj: Self-supervised Foundation Models as Rewards for Label-free 3D Object Segmentationcited by 0
  17. 2026ESAM++: Efficient Online 3D Perception on the Edgecited by 0
  18. 2026SAD-GS: Learning Reliable 3D Semantic Gaussian Fields via Dynamic Geo-Semantic Anchoringcited by 0
  19. 2026CDIS: Cross-Dimensional Class-Agnostic 3D Instance Segmentation via 2D Mask Tracking and 3D-2D Projection Mergingcited by 0