Research map / Detection and segmentation / 3D detection and driving
Open-Vocabulary 3D Object Detection with Co-Distillation Discovery and Dual Guidance Robust Training
Shangbo Yuan, Jie Xu, Xiaofeng Zhu, Na Zhao
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 SGUR (scene-awareness-guided uncertainty regularization) (ablation) | Co-3DGT · SUN RGB-D | AP25 mean: 20.62 -> 22.17 | Improved |
| add SGUR (ablation) | Co-3DGT · ScanNetV2 | AP25 mean: 20.15 -> 23.12 | Improved |
| add LGHA (LLM-guided hierarchical alignment) (ablation) | Co-3DGT · SUN RGB-D | AP25 mean: 20.62 -> 21.35 | Improved |
| add LGHA (ablation) | Co-3DGT · ScanNetV2 | AP25 mean: 20.15 -> 22.73 | Improved |
| combine SGUR and LGHA (ablation) | Co-3DGT · SUN RGB-D | AP25 mean: 20.62 -> 22.63 | Improved |
| combine SGUR and LGHA (ablation) | Co-3DGT · ScanNetV2 | AP25 mean: 20.15 -> 23.75 | Improved |
| use Co-distil(CuTR) discovery instead of 3D-Proposal-based discovery (ablation) | novel object discovery · SUN RGB-D | AP25 novel (discovery): 8.49 -> 10.97 | Improved |
| use Co-distil(CuTR) instead of 2D-Detection-based discovery for final training (ablation) | detector without dual guidance · ScanNetV2 | AP25 novel (detection): 11.92 -> 18.74 | Improved |
| add pretrained CuTR class-agnostic 3D detector to co-distillation from scratch (ablation) | Co-distillation · SUN RGB-D | AP25 novel (discovery): 10.45 -> 10.97 | Improved |
| swap ChatGPT-5 for Llama-3.3-70B-Instruct as hierarchy generator (ablation) | LGHA · SUN RGB-D | AP25 mean: 22.63 -> 21.42 | No clear effect |
| swap ChatGPT-5 for Gemini-3-Flash-Thinking (ablation) | LGHA · ScanNetV2 | AP25 mean: 23.75 -> 23.84 | No clear effect |
| replace OV-Uni3DETR with Co-3DGT | Co-3DGT · ScanNetV2 | AP25 novel: 12.09 -> 21.91 | 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.
- OV-Uni3DETR — OV-Uni3DETR: Towards Unified Open-Vocabulary 3D Object Detection via Cycle-Modality Propagation (ECCV 2024) · better on 7 of 7 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
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.90)
- FM-OV3D: Foundation Model-Based Cross-Modal Knowledge Blending for Open-Vocabulary 3D Detection (AAAI 2024 · similarity 0.89)
- Towards 3D Objectness Learning in an Open World (NeurIPS 2025 · similarity 0.89)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.89)
- Hierarchical Cross-Modal Alignment for Open-Vocabulary 3D Object Detection (AAAI 2025 · similarity 0.89)
- Open-Vocabulary 3D Semantic Segmentation with Foundation Models (CVPR 2024 · 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.
