mlsanityData to 7 Oct 2026

PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding

arXiv 2510.20155 ↗First posted 23 Oct 2025cs.CVcited by 24

Not on the map. Why?

Abstract (arXiv)

Understanding objects at the level of their constituent parts is fundamental to advancing computer vision, graphics, and robotics. While datasets like PartNet have driven progress in 3D part understanding, their reliance on untextured geometries and expert-dependent annotation limits scalability and usability. We introduce PartNeXt, a next-generation dataset addressing these gaps with over 23,000 high-quality, textured 3D models annotated with fine-grained, hierarchical part labels across 50 categories. We benchmark PartNeXt on two tasks: (1) class-agnostic part segmentation, where state-of-the-art methods (e.g., PartField, SAMPart3D) struggle with fine-grained and leaf-level parts, and (2) 3D part-centric question answering, a new benchmark for 3D-LLMs that reveals significant gaps in open-vocabulary part grounding. Additionally, training Point-SAM on PartNeXt yields substantial gains over PartNet, underscoring the dataset's superior quality and diversity. By combining scalable annotation, texture-aware labels, and multi-task evaluation, PartNeXt opens new avenues for research in structured 3D understanding.

Key facts. Cited by 24 papers in the map (data through 7 Oct 2026). Builds on 48 of its 57 references that are in the map. First posted on arXiv 23 Oct 2025.

Builds on

48 of 57 references.

References · 48

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2025Qwen3 Technical Reportcited by 8,195
  3. 2023Segment Anythingcited by 8,158
  4. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  5. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,223
  6. 2024GPT-4o System Cardcited by 4,467
  7. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,384
  8. 2022Objaverse: A Universe of Annotated 3D Objectscited by 1,522
  9. 2016A Point Set Generation Network for 3D Object Reconstruction from a Single Imagecited by 972
  10. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 878
  11. 2023Objaverse-XL: A Universe of 10M+ 3D Objectscited by 767
  12. 2021Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstructioncited by 646
36 more
  1. 2020SAPIEN: A SimulAted Part-based Interactive ENvironmentcited by 584
  2. 2019Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Datacited by 532
  3. 2018PartNet: A Large-scale Benchmark for Fine-grained and Hierarchical Part-level 3D Object Understandingcited by 521
  4. 2022GET3D: A Generative Model of High Quality 3D Textured Shapes Learned from Imagescited by 446
  5. 2021ABO: Dataset and Benchmarks for Real-World 3D Object Understandingcited by 360
  6. 2023PointLLM: Empowering Large Language Models to Understand Point Cloudscited by 326
  7. 2023OmniObject3D: Large-Vocabulary 3D Object Dataset for Realistic Perception, Reconstruction and Generationcited by 324
  8. 20203D-FUTURE: 3D Furniture shape with TextUREcited by 306
  9. 2023MVImgNet: A Large-scale Dataset of Multi-view Imagescited by 219
  10. 2023Scalable 3D Captioning with Pretrained Modelscited by 201
  11. 2016Thingi10K: A Dataset of 10,000 3D-Printing Modelscited by 182
  12. 2022Ditto: Building Digital Twins of Articulated Objects from Interactioncited by 152
  13. 2022GAPartNet: Cross-Category Domain-Generalizable Object Perception and Manipulation via Generalizable and Actionable Partscited by 135
  14. 2024ShapeLLM: Universal 3D Object Understanding for Embodied Interactioncited by 133
  15. 2022PartSLIP: Low-Shot Part Segmentation for 3D Point Clouds via Pretrained Image-Language Modelscited by 117
  16. 2023PARIS: Part-level Reconstruction and Motion Analysis for Articulated Objectscited by 115
  17. 2024URDFormer: A Pipeline for Constructing Articulated Simulation Environments from Real-World Imagescited by 106
  18. 20213D AffordanceNet: A Benchmark for Visual Object Affordance Understandingcited by 100
  19. 2024SAMPart3D: Segment Any Part in 3D Objectscited by 95
  20. 2025PARTFIELD: Learning 3D Feature Fields for Part Segmentation and Beyondcited by 91
  21. 2024Articulate-Anything: Automatic Modeling of Articulated Objects via a Vision-Language Foundation Modelcited by 90
  22. 2022AKB-48: A Real-World Articulated Object Knowledge Basecited by 81
  23. 2023GPT4Point: A Unified Framework for Point-Language Understanding and Generationcited by 69
  24. 2024SINGAPO: Single Image Controlled Generation of Articulated Parts in Objectscited by 66
  25. 2020Generative 3D Part Assembly via Dynamic Graph Learningcited by 65
  26. 2024Real2Code: Reconstruct Articulated Objects via Code Generationcited by 64
  27. 2020Learning 3D Part Assembly from a Single Imagecited by 51
  28. 2024Point-SAM: Promptable 3D Segmentation Model for Point Cloudscited by 49
  29. 2023PartSLIP++: Enhancing Low-Shot 3D Part Segmentation via Multi-View Instance Segmentation and Maximum Likelihood Estimationcited by 37
  30. 2024Segment Any Meshcited by 37
  31. 2024AffordDP: Generalizable Diffusion Policy with Transferable Affordancecited by 35
  32. 2020Learning to Group: A Bottom-Up Framework for 3D Part Discovery in Unseen Categoriescited by 29
  33. 2023Category-Level Multi-Part Multi-Joint 3D Shape Assemblycited by 22
  34. 2024SeqAfford: Sequential 3D Affordance Reasoning via Multimodal Large Language Modelcited by 18
  35. 2025Infinite Mobility: Scalable High-Fidelity Synthesis of Articulated Objects via Procedural Generationcited by 17
  36. 2024Part-Guided 3D RL for Sim2Real Articulated Object Manipulationcited by 9

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 22.

24 papers.

Sort:

By area:

  1. 2026HY3D-Bench: Generation of 3D Assetscited by 11
  2. 2026LaS-Comp: Zero-shot 3D Completion with Latent-Spatial Consistencycited by 9
  3. 2026SegviGen: Repurposing 3D Generative Model for Part Segmentationcited by 9
  4. 2025Kinematify: Open-Vocabulary Synthesis of High-DoF Articulated Objectscited by 8
  5. 2026ArtLLM: Generating Articulated Assets via 3D LLMcited by 8
  6. 2026Feedforward 3D Editing Learns from Semantic-Part Transformationcited by 7
  7. 2026Reconstruction by Generation: 3D Multi-Object Scene Reconstruction from Sparse Observationscited by 5
  8. 2026Artiverse: A Diverse and Physically Grounded Dataset for Articulated Objectscited by 4
  9. 2026VideoAfford: Grounding 3D Affordance from Human-Object-Interaction Videos via Multimodal Large Language Modelcited by 2
  10. 2026AssemLM: A Spatial Reasoning Multimodal Large Language Model for Robotic Assemblycited by 2
  11. 2026Instruct-Particulate: Scaling Feed-Forward 3D Object Articulation with Kinematic Controlcited by 2
  12. 2026Sequential Planning via Anchored Robotic Keypointscited by 2
12 more
  1. 2026R3D: Revisiting 3D Policy Learningcited by 1
  2. 2026PartLLM: A Unified Multimodal Foundation for 3D Part Segmentationcited by 1
  3. 2025XDen-1K: A Density Field Dataset of Real-World Objectscited by 0
  4. 2026CRAG: Can 3D Generative Models Help 3D Assembly?cited by 0
  5. 2026PARSE: Part-Aware Relational Spatial Modelingcited by 0
  6. 2026Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editingcited by 0
  7. 2026MegaParts: Scaling Part-Aware 3D Object Generation to 300 Parts via Token-Efficient Autoregressive Modelingcited by 0
  8. 2026KaiNinja: Extending Native 3D Generators to the Part Levelcited by 0
  9. 2026PART: Learning 3D Part Assembly and Retrieval with Transformerscited by 0
  10. 2026Scaling Versatile 3D Assets Editing with a Million-Scale Datasetcited by 0
  11. 2026DirectUV: Image-Conditioned UV Texture Generation with Surface-Aware Positional Encodingcited by 0
  12. 2026Point2Part: Unified 3D Partitioning from Point Promptscited by 0