Research map / Vision-language and multimodal / CLIP and zero-shot
OpenGround: Planning-based Online Perception for Open-World 3D Visual Grounding
Wenyuan Huang, zhenyu zhang, Zhao Wang, Zhou Wei, Ting Huang, Fang Zhao, Jian Yang
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 Task-Chain Planning order instead of Jump/Random/Difficulty/Relevance orders (ablation) | OpenGround · OpenTarget | Acc@0.50: 29.2 (Random) / 29.8 (Jump) / 31.5 (Difficulty) / 32.6 (Relevance) -> 34.2 | Improved |
| provide initial Mask3D OLT as prior (ablation) | OpenGround · OpenTarget | Acc@0.50: 27.1 -> 34.2 | Improved |
| use smaller VLM (Qwen3-VL-32B vs GLM-4.5V) (ablation) | OpenGround · OpenTarget | Acc@0.50: 34.2 -> 30.4 | Hurt |
| use context-guided view selection instead of SeeGround/SeqVLM/Struct2D views (ablation) | OpenGround single-step grounding · OpenTarget | Acc@0.50: 31.6 / 31.8 / 26.3 -> 34.2 | Improved |
| annotate previously grounded objects plus candidates instead of all candidates or all mentioned objects (ablation) | OpenGround · OpenTarget | Acc@0.50: 26.5 / 25.3 -> 34.2 | Improved |
| add structured query decomposition in VLM reasoning (ablation) | OpenGround · OpenTarget | Acc@0.50: 32.9 -> 34.2 | Improved |
| provide original non-annotated images to VLM (ablation) | OpenGround · OpenTarget | Acc@0.50: 33.6 -> 34.2 | Improved |
| maximize coverage of context objects in view selection (ablation) | OpenGround · OpenTarget | Acc | Improved |
| increase views V from 3 to 4 (ablation) | OpenGround CGP · OpenTarget | Acc: 0.52% gain, +33.3% views | No clear effect |
| use OpenGround with SPAZER backend (ablation) | OpenGround · Nr3D | accuracy: 63.8 -> 64.8 | Improved |
| OpenGround vs SPAZER with GPT-4o | OpenGround · Nr3D | accuracy: 63.8 -> 61.7 | Hurt |
| apply OpenGround to standard closed-set benchmark | OpenGround vs SeqVLM (Doubao) · ScanRefer | Acc@0.50: 49.6 -> 54.4 | Improved |
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.
- SeeGround — SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025) · better on 4 of 4 shared results
- SPAZER — SPAZER: Spatial-Semantic Progressive Reasoning Agent for Zero-shot 3D Visual Grounding (NeurIPS 2025) · 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
- SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025 · similarity 0.91)
- From Objects to Anywhere: A Holistic Benchmark for Multi-level Visual Grounding in 3D Scenes (NeurIPS 2025 · similarity 0.91)
- Visual Programming for Zero-shot Open-Vocabulary 3D Visual Grounding (CVPR 2024 · similarity 0.90)
- CoT3DRef: Chain-of-Thoughts Data-Efficient 3D Visual Grounding (ICLR 2024 · similarity 0.90)
- DenseGrounding: Improving Dense Language-Vision Semantics for Ego-centric 3D Visual Grounding (ICLR 2025 · similarity 0.90)
- Grounded 3D-Aware Spatial Vision-Language Modeling (CVPR 2026 · similarity 0.90)
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.
