Research map / Point clouds / Segmentation and detection
3DOS: Towards 3D Open Set Learning - Benchmarking and Understanding Semantic Novelty Detection on Point Clouds
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the learned feature embedding with nearest-training-sample L2 distance rather than MSP | DGCNN on Synthetic Benchmark SN1 · Synthetic Benchmark SN1 | AUROC: 74.0 -> 80.4 | Helped |
| Replace MSP with GradNorm | 3D discriminative methods · Synthetic Benchmark | AUROC: 85.2 -> 72.9 | Hurt |
| Use ReAct instead of MSP | DGCNN · Synthetic Benchmark average | AUROC: 85.2 -> 88.4 | Helped |
| Train a normalizing flow on the DGCNN feature embedding instead of using MSP | DGCNN · Synthetic to Real average | AUROC: 66.7 -> 71.3 | Helped |
| Use OE+mixup finetuning instead of MSP | DGCNN · Synthetic to Real average | AUROC: 66.7 -> 65.3 | Hurt |
| Add label smoothing (ablation) | DGCNN classifier · Synthetic Benchmark SN1 | AUROC: 74.0 -> 77.4 | Helped |
| Add label smoothing to the CE (L2) model (ablation) | PointNet++ classifier · Synthetic Benchmark SN1 | AUROC: 83.4 -> 79.1 | Hurt |
| Increase the number of sampled points (ablation) | CE (L2) · Synthetic Benchmark SN1 | AUROC | No clear effect |
| Add Occlusion and LIDAR real-world augmentations (ablation) | MSP with DGCNN · Synthetic to Real average | AUROC: 66.7 -> 73.9 | Helped |
| Add Occlusion and LIDAR real-world augmentations (ablation) | SubArcFace with DGCNN · Synthetic to Real average | AUROC: 71.6 -> 75.1 | Helped |
| Add Occlusion and LIDAR real-world augmentations (ablation) | MSP with PointNet++ · Synthetic to Real average | AUROC: 75.6 -> 75.5 | No clear effect |
| Match test point count to training point count at 1024 (ablation) | Synthetic-to-real models · Synthetic to Real average | AUROC | Mixed |
Closest papers by meaning
- Towards 3D Objectness Learning in an Open World (NeurIPS 2025 · similarity 0.88)
- OpenNeRF: Open Set 3D Neural Scene Segmentation with Pixel-Wise Features and Rendered Novel Views (ICLR 2024 · similarity 0.88)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.87)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.87)
- MESC-3D:Mining Effective Semantic Cues for 3D Reconstruction from a Single Image (CVPR 2025 · similarity 0.87)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.86)
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.