Research map / Detection and segmentation / 3D detection and driving
FoundObj: Self-supervised Foundation Models as Rewards for Label-free 3D Object Segmentation
Zihui Zhang, Zhixuan Sun, Yafei YANG, Jinxi Li, Jiahao Chen, Bo Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add geometric reward module (ablation) | FoundObj · ScanNet val | AP: 19.5 -> 24.2 | Improved |
| add semantic reward module (ablation) | FoundObj · ScanNet val | AP: 15.3 -> 24.2 | Improved |
| DBSCAN radius 0.05 instead of 0.02 (ablation) | FoundObj · ScanNet val | AP: 21.9 -> 24.2 | Improved |
| DBSCAN radius 0.05 instead of 0.1 (ablation) | FoundObj · ScanNet val | AP: 22.5 -> 24.2 | Improved |
| point ratio 30% instead of 20% (ablation) | FoundObj · ScanNet val | AP: 22.1 -> 24.2 | Improved |
| neighbor threshold 0.1 instead of 0.2 (ablation) | FoundObj · ScanNet val | AP: 21.4 -> 24.2 | Improved |
| cost bank size 20 instead of 10 (ablation) | FoundObj · ScanNet val | AP: 20.9 -> 24.2 | Improved |
| cost bank size 20 instead of 40 (ablation) | FoundObj · ScanNet val | AP: 21.1 -> 24.2 | Improved |
| neighbor threshold 0.1 instead of 0.05 (ablation) | FoundObj · ScanNet val | AP: 23.0 -> 24.2 | Improved |
| use RL agent with TRELLIS+DINO features instead of UnScene3D clustering (ablation) | FoundObj · ScanNet val, DINO+TRELLIS | AP: 15.3 -> 24.2 | Improved |
| use RL agent instead of Part2Object on DINO-only features (ablation) | FoundObj · ScanNet val, DINO-only | AP: 19.6 -> 19.5 | No clear effect |
| clean pseudo masks using ground-truth IoU filtering (ablation) | Mask3D training · ScanNet val | AP: 24.2 -> 37.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- UnScene3D — UnScene3D: Unsupervised 3D Instance Segmentation for Indoor Scenes (CVPR 2024) · better on 5 of 5 shared results
- EFEM — EFEM: Equivariant Neural Field Expectation Maximization for 3D Object Segmentation Without Scene Supervision (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GrabS: Generative Embodied Agent for 3D Object Segmentation without Scene Supervision (ICLR 2025 · similarity 0.88)
- LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds (CVPR 2025 · similarity 0.87)
- Point-SAM: Promptable 3D Segmentation Model for Point Clouds (ICLR 2025 · similarity 0.87)
- AGILE3D: Attention Guided Interactive Multi-object 3D Segmentation (ICLR 2024 · similarity 0.87)
- Weakly Supervised Point Cloud Semantic Segmentation via Artificial Oracle (CVPR 2024 · similarity 0.87)
- Relation3D : Enhancing Relation Modeling for Point Cloud Instance Segmentation (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
