Research map / 3D vision / 3D generation
Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional Conditioning
Changbo Yan, Zhongbo Zhang, Zaibin Zhang, Yifan Wang, Lijun Wang, Huchuan Lu
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 intra-target saliency and backgroundness to targetness-only (ablation) | Attention-DP3 (T only) · Door/Stick-Pull/Pick-Place/Push-Wall avg | success rate: 0.5425 -> 0.72 | Improved |
| add targetness to intra-target saliency plus backgroundness (ablation) | Attention-DP3 (B+I) · Door/Stick-Pull/Pick-Place/Push-Wall avg | success rate: 0.595 -> 0.72 | Improved |
| add intra-target saliency to targetness plus backgroundness (ablation) | Attention-DP3 (T+B) · Door/Stick-Pull/Pick-Place/Push-Wall avg | success rate: 0.5625 -> 0.72 | Improved |
| add backgroundness to targetness plus saliency (ablation) | Attention-DP3 (T+I) · Door/Stick-Pull/Pick-Place/Push-Wall avg | success rate: 0.5175 -> 0.72 | Improved |
| encode each field with its own AttnEncoder instead of a single encoder (ablation) | Attention-DP3 · Door/Stick-Pull/Pick-Place/Push-Wall avg | success rate: 0.5225 -> 0.72 | Improved |
| use late fusion instead of early fusion (ablation) | Attention-DP3 · Door/Stick-Pull/Pick-Place/Push-Wall avg | success rate: 0.255 -> 0.72 | Improved |
| add Tri-Field Attentional Conditioning to DP3 | DP3 · MetaWorld overall | success rate: 0.669 -> 0.726 | Improved |
| add Tri-Field Attentional Conditioning to DP3 | DP3 · Adroit avg | success rate: 0.68 -> 0.78 | Improved |
| add Tri-Field Attentional Conditioning to DP3 | DP3 · DexArt avg | success rate: 0.52 -> 0.56 | Improved |
| add Tri-Field Attentional Conditioning to DP3 | DP3 · SO101 real world avg | success rate: 0.52 -> 0.73 | Improved |
| add Tri-Field Attentional Conditioning under extra real-world distractors | DP3 · SO101 +Extra distractors | success rate: 0.13 -> 0.45 | Improved |
| replace predicted masks with ground-truth masks (ablation) | Attention-DP3 · Mask perturbation avg | success rate: 0.524 -> 0.542 | No clear effect |
Closest papers by meaning
- CTRL&SHIFT: High-quality Geometry-Aware Object Manipulation in Visual Generation (ICLR 2026 · similarity 0.88)
- 3DP3: 3D Scene Perception via Probabilistic Programming (NeurIPS 2021 · similarity 0.87)
- DORSal: Diffusion for Object-centric Representations of Scenes $\textit{et al.}$ (ICLR 2024 · similarity 0.87)
- Promptable 3-D Object Localization with Latent Diffusion Models (NeurIPS 2025 · similarity 0.87)
- 3DiffTection: 3D Object Detection with Geometry-Aware Diffusion Features (CVPR 2024 · similarity 0.87)
- Diffusion Models are Geometry Critics: Single Image 3D Editing Using Pre-Trained Diffusion Priors (ECCV 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.
