Research map / 3D vision / Face and shape
TouchAnything: Diffusion-Guided 3D Reconstruction from Sparse Robot Touches
Langzhe Gu, Hung-Jui Huang, Mohamad Qadri, Michael Kaess, Wenzhen Yuan
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 class-level text prompt instead of incorrect prompt (an airplane) (ablation) | TouchAnything · Real objects, 20 touches (4 objects) | EMD: bottle 0.0105 -> 0.0057; camera 0.0180 -> 0.0098 | Improved |
| use class-level prompt instead of empty prompt (ablation) | TouchAnything · Real objects, 20 touches (4 objects) | EMD: potted meat can 0.0086 -> 0.0062; tomato soup can 0.0089 -> 0.0077 | Improved |
| use detailed instance-specific prompt instead of class-level prompt (ablation) | TouchAnything · Real objects, 20 touches | EMD: bottle 0.0057 -> 0.0089; camera 0.0098 -> 0.0097 | No clear effect |
| use all available touches instead of 20 or 40 (ablation) | TouchAnything · Real objects | EMD: bottle p3 0.0057 -> 0.0066 (all) | No clear effect |
| add diffusion prior (SDS) to tactile-only reconstruction (ablation) | TouchAnything · Real objects | reconstruction quality (qualitative) | Improved |
| add stage 2 DMTet refinement at 512x512 rendering (ablation) | TouchAnything · Real objects | surface detail (qualitative) | Improved |
| replace trained tactile priors with pretrained 2D diffusion guidance | TouchAnything vs TouchSDF · ShapeNet Guitar (20 touches) | EMD: 0.155 -> 0.060 | Improved |
| replace Touch2Shape with TouchAnything | TouchAnything vs Touch2Shape · ShapeNet Camera (20 touches) | EMD: 0.056 -> 0.047 | Improved |
| replace Touch2Shape with TouchAnything | TouchAnything vs Touch2Shape · ShapeNet Jar (20 touches) | EMD: 0.055 -> 0.053 | 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.
- Touch2Shape — Touch2Shape: Touch-Conditioned 3D Diffusion for Shape Exploration and Reconstruction (CVPR 2025) · better on 6 of 6 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
- Touch2Shape: Touch-Conditioned 3D Diffusion for Shape Exploration and Reconstruction (CVPR 2025 · similarity 0.90)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.87)
- Tactile-Augmented Radiance Fields (CVPR 2024 · similarity 0.87)
- EgoTac: In-the-wild Tactile Prediction from Egocentric Vision (NeurIPS 2026 · similarity 0.87)
- Tactile DreamFusion: Exploiting Tactile Sensing for 3D Generation (NeurIPS 2024 · similarity 0.86)
- AnyTouch: Learning Unified Static-Dynamic Representation across Multiple Visuo-tactile Sensors (ICLR 2025 · similarity 0.86)
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.
