Research map / Detection and segmentation / 3D detection and driving
OV-SCAN: Semantically Consistent Alignment for Novel Object Discovery in Open-Vocabulary 3D Object Detection
Adrian Chow, Evelien Riddell, Yimu Wang, Sean Sedwards, Krzysztof Czarnecki
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 |
|---|---|---|---|
| use adaptive 3D box search instead of greedy search at same 150,000 iterations (ablation) | OV-SCAN SC-NOD · nuScenes | mAP: 25.4 -> 31.1 | Improved |
| use adaptive search with only 37,500 iterations instead of 150,000 greedy (ablation) | OV-SCAN SC-NOD · nuScenes | mAP: 25.4 -> 28.9 | Improved |
| increase adaptive search iterations from 37,500 to 150,000 (ablation) | OV-SCAN SC-NOD · nuScenes | mAP: 28.9 -> 31.1 | Improved |
| balance cost weights (5.0,1.0,1.0,3.0) instead of density-only (ablation) | OV-SCAN SC-NOD · nuScenes | mAP: 26.1 -> 31.1 | Improved |
| use tuned weights instead of equal weights (ablation) | OV-SCAN SC-NOD · nuScenes | mAP: 27.4 -> 31.1 | Improved |
| add resolution filter (threshold 4000) (ablation) | OV-SCAN selective alignment · nuScenes | mAP: 30.4 -> 31.1 | Improved |
| raise resolution filter threshold from 4000 to 10000 (ablation) | OV-SCAN selective alignment · nuScenes | mAP: 31.1 -> 30.6 | Hurt |
| add fixed occlusion filter (ablation) | OV-SCAN selective alignment · nuScenes | mAP: 29.4 -> 30.7 | Improved |
| use class-based occlusion thresholds instead of fixed (ablation) | OV-SCAN selective alignment · nuScenes | mAP: 30.7 -> 31.1 | Improved |
| add two-stage class-guided alignment to one-stage FFN (ablation) | OV-SCAN alignment head · nuScenes | mAP: 25.2 -> 29.4 | Improved |
| add text-prompt guided alignment (H2SA) to two-stage FFN (ablation) | OV-SCAN alignment head · nuScenes | mAP: 29.4 -> 31.1 | Improved |
| add camera image backbone and fine-tune (OV-SCAN-Fusion) | OV-SCAN · nuScenes | mAP: 31.1 -> 33.8 | 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.
- OpenSight — OpenSight: A Simple Open-Vocabulary Framework for LiDAR-Based Object Detection (ECCV 2024) · better on 2 of 2 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
- Hierarchical Cross-Modal Alignment for Open-Vocabulary 3D Object Detection (AAAI 2025 · similarity 0.92)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.90)
- Open-Vocabulary 3D Semantic Segmentation with Foundation Models (CVPR 2024 · similarity 0.90)
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.89)
- FM-OV3D: Foundation Model-Based Cross-Modal Knowledge Blending for Open-Vocabulary 3D Detection (AAAI 2024 · similarity 0.88)
- CSV-Occ: Fusing Multi-frame Alignment for Occupancy Prediction with Temporal Cross State Space Model and Central Voting Mechanism (ICML 2025 · similarity 0.88)
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.
