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

TL;DR

<3-5 sentence high-level summary> PartNeXt presents a large-scale, textured 3D part dataset with fine-grained, hierarchical annotations across 50 categories to overcome PartNet's texture and annotation scalability limitations. It introduces a scalable web-based annotation system combining hierarchical taxonomies generated with AI tools (GPT-4o, CLIP) and per-face labeling on textured meshes, enabling rich, texture-aware part labels. The paper benchmarks two tasks—class-agnostic 3D part segmentation and 3D part-centric question answering—and evaluates open-vocabulary grounding with 3D LLMs, revealing substantial gaps. Experiments show that training Point-SAM on PartNeXt yields clear gains over PartNet, underscoring PartNeXt’s value as a foundation for next-generation 3D part understanding and grounding research.

Abstract

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.

Figures (20)

• Figure 1: We present PartNeXt, a next-generation dataset tailored for fine-grained, hierarchically structured 3D part understanding.
• Figure 2: Illustration of our annotation interface. The example shows a microwave containing an internal tray. The dual-panel layout allows annotators to first label external parts such as the "door" (as shown in the right panel with already segmented meshes), and then proceed to annotate internal components like the "tray" (visible in the unsegmented mesh in the left panel). This design effectively mitigates occlusion issues during annotation.
• Figure 3: PartNeXt dataset. We visualize example shapes with fine-grained part annotations for the 50 object categories in PartNeXt.
• Figure 4: Part Segmentation Results on PartNeXt. PartField struggles to separate connected regions, SAMesh excels at fine-grained segmentation but over-segments, while SAMPart3D lacks continuity in weak textures and granularity control.
• Figure 5: Representative prompt–response pairs used to evaluate 3D part-level understanding. (a) Part Counting: the model is requested to enumerate the number of legs in a chair. (b) Part Classification: the model must name the part highlighted in red within the point-cloud bed. (c) Part Grounding: the model is asked to localize the “Shelf” of a bookcase by outputting the eight corner coordinates of its bounding box.