Research map / 3D vision / 3D generation
NegROI: Click-Centric Uncertainty-Guided Refinement with Scene-Conditioned Negative Prompts for Robust Interactive 3D Segmentation
shuheng zhang, Feng Wu
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 scene-conditioned negative prompts (ablation) | NegROI base · ScanNet40 | IoU@1: 67.0 -> 69.2 | Improved |
| add scene-conditioned negative prompts (ablation) | NegROI base · S3DIS | IoU@1: 62.5 -> 64.2 | Improved |
| add click-centric fine-grid ROI refinement (ablation) | NegROI + NEG prompts · ScanNet40 | IoU@1: 69.2 -> 70.9 | Improved |
| add click-centric fine-grid ROI refinement (ablation) | NegROI + NEG prompts · S3DIS | IoU@1: 64.2 -> 65.8 | Improved |
| add uncertainty-driven ROI selection (ablation) | NegROI + ROI · ScanNet40 | IoU@1: 70.9 -> 71.8 | Improved |
| add uncertainty-driven ROI selection (ablation) | NegROI + ROI · ScanNet40 | IoU@10: 83.0 -> 82.8 | Hurt |
| add boundary hard negative loss L_hn (ablation) | NegROI + uncertainty ROI · ScanNet40 | IoU@1: 71.8 -> 71.2 | Hurt |
| add boundary hard negative loss L_hn (ablation) | NegROI + uncertainty ROI · S3DIS | IoU@10: 87.4 -> 87.1 | Hurt |
| add diversity regularizer L_div (ablation) | NegROI + L_hn · ScanNet40 | IoU@1: 71.2 -> 72.1 | Improved |
| add diversity regularizer L_div (ablation) | NegROI + L_hn · S3DIS | IoU@1: 65.6 -> 66.9 | Improved |
| add full NegROI (all components) over base (ablation) | NegROI · ScanNet40 | IoU@10: 80.8 -> 83.6 | Improved |
| use NegROI instead of Easy3D, out-of-domain to KITTI-360 | NegROI · KITTI-360 (train ScanNet40) | IoU@10: 83.6 -> 87.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Easy3D — Easy3D: A Simple Yet Effective Method for 3D Interactive Segmentation (ICCV 2025) · better on 7 of 8 shared results
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
- Probabilistic Interactive 3D Segmentation with Hierarchical Neural Processes (ICML 2025 · similarity 0.89)
- AGILE3D: Attention Guided Interactive Multi-object 3D Segmentation (ICLR 2024 · similarity 0.89)
- NTClick: Achieving Precise Interactive Segmentation With Noise-tolerant Clicks (CVPR 2025 · similarity 0.87)
- Order-aware Interactive Segmentation (ICLR 2025 · similarity 0.86)
- Rethinking Interactive Image Segmentation with Low Latency High Quality and Diverse Prompts (CVPR 2024 · similarity 0.86)
- Prompting Diffusion Models for Zero-Shot Instance Segmentation (NeurIPS 2026 · similarity 0.85)
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.
