Research map / Detection and segmentation / 3D detection and driving
Zoo3D: Zero-Shot 3D Object Detection at Scene Level
Andrey Lemeshko, Bulat Gabdullin, Nikita Drozdov, Anton Konushin, Danila Rukhovich, Maksim Kolodiazhnyi
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 |
|---|---|---|---|
| train class-agnostic TR3D detector on pseudo labels (Zoo3D_1 vs Zoo3D_0) | Zoo3D · Point cloud + posed images, ScanNet20 | mAP25: 34.7 -> 37.2 | Improved |
| second iteration of pseudo-label retraining (Zoo3D_2) (ablation) | Zoo3D_1 · Class-agnostic ScanNet200, posed images | mAP25: 36.1 -> 37.6 | Improved |
| first iteration of self-training (ablation) | Zoo3D_0 · Class-agnostic ScanNet200, posed images | mAP25: 22.4 -> 36.1 | Improved |
| add occlusion filter to open-vocabulary module (ablation) | Zoo3D_0 · Open-vocabulary module ablation, posed images ScanNet200 | mAP25: 14.7 -> 14.8 | No clear effect |
| add SAM mask refinement (ablation) | Zoo3D_0 · Open-vocabulary module ablation, posed images ScanNet200 | mAP25: 14.8 -> 15.4 | Improved |
| add multi-scale processing (ablation) | Zoo3D_0 · Open-vocabulary module ablation, posed images ScanNet200 | mAP50: 5.7 -> 6.3 | Improved |
| add multi-scale processing (ablation) | Zoo3D_0 · Open-vocabulary module ablation, posed images ScanNet200 | mAP25: 15.4 -> 16.5 | Improved |
| increase number of images per scene from 15 to 45 (ablation) | Zoo3D_0 · Posed images ScanNet200, number of views | mAP25: 9.6 -> 14.3 | Improved |
| use DUSt3R instead of DROID-SLAM for reconstruction (ablation) | Zoo3D_1 · Unposed images ScanNet200 class-agnostic | mAP25: 2.4 -> 19.0 | Improved |
| use VGGT instead of DUSt3R for reconstruction (ablation) | Zoo3D_1 · Unposed images ScanNet200 class-agnostic | mAP25: 19.0 -> 28.2 | Improved |
| assign all objects to 16-cm level instead of 32-cm level (ablation) | Zoo3D_1 class-agnostic detector · Posed images ScanNet200 class-agnostic | mAP25: 31.2 -> 36.1 | Improved |
| assign objects 50/50 by size instead of all to 16-cm level (ablation) | Zoo3D_1 class-agnostic detector · Posed images ScanNet200 class-agnostic | mAP25: 36.1 -> 34.5 | Hurt |
Later papers that compared with this method
1 later accepted paper reports Zoo3D_1 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Group3D: MLLM-Guided Semantic Grouping for Open-Vocabulary 3D Object Detection (ECCV 2026 · better on 12 of 12 shared results)
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.
- OpenM3D — OpenM3D: Open Vocabulary Multi-view Indoor 3D Object Detection without Human Annotations (ICCV 2025) · better on 3 of 3 shared results
- OV-Uni3DETR — OV-Uni3DETR: Towards Unified Open-Vocabulary 3D Object Detection via Cycle-Modality Propagation (ECCV 2024) · better on 3 of 3 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
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.91)
- LabelAny3D: Label Any Object 3D in the Wild (NeurIPS 2025 · similarity 0.91)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.89)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.89)
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.88)
- Towards 3D Objectness Learning in an Open World (NeurIPS 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.
