AI Paper Atlas

Research map / 3D vision / 3D generation

H2OFlow: Grounding Human-Object Affordances with 3D Generative Models and Dense Diffused Flows

Harry Zhang, Luca Carlone

ICLR 2026 · 3D vision › 3D generation · approach: Hand-object interaction generation

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 changedWhereMeasuredEffect
learn dense diffused flows instead of direct SMPL parameters (ablation)H2OFlow · OMOMOSIM-H (%): 57.3 -> 72.3Improved
add cross-attention weights to affordance aggregation (ablation)H2OFlow · OMOMOSIM-H (%): 67.3 -> 72.3Improved
add orientational affordance to contact (ablation)H2OFlow · HOI region retrievalmAP@10: 0.66 -> 0.69Improved
add spatial affordance to contact (ablation)H2OFlow · HOI region retrievalmAP@10: 0.66 -> 0.68Improved
use all three affordances (contact, orientation, spatial) (ablation)H2OFlow · HOI region retrievalmAP@10: 0.66 -> 0.76Improved
shuffle orientational or spatial affordances (ablation)H2OFlow · HOI pose selectionTop-5 accuracy: 0.64 -> 0.44Hurt
occlude 50% of the object point cloud (ablation)H2OFlow · OMOMOSIM-H (%): 72.3 -> 70.9No clear effect
reconstruct mesh from point cloud for COMA input (ablation)COMA · OMOMOSIM-H (%): 41.3 -> 20.7Hurt
fine-tune H2OFlow on BEHAVE objects (ablation)H2OFlow · BEHAVESIM-H (%): 55.3 -> 74.4Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy