atlas

Research map / 3D vision / 3D generation

OAKINK2: A Dataset of Bimanual Hands-Object Manipulation in Complex Task Completion

CVPR 2024 · 3D vision › 3D generation · approach: Hand-object interaction generation · 24 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Introduces OAKINK2, a multi-view bimanual hand-object manipulation dataset organized into affordance, primitive, and complex-task levels, and a Complex Task Completion framework using GPT-4 planning, oracle object retargeting, and a diffusion-based hand motion generator (MF-MDM).

Problem

Human-level understanding of bimanual object manipulation for long-horizon complex tasks lacks datasets that show both how affordances are fulfilled and how sub-actions depend on each other.

Approach

An object-centric three-level abstraction (affordance, primitive task, complex task with a primitive dependency graph) structures real mocap demonstrations, and a pipeline uses an LLM to decompose complex tasks into primitives and a task-aware diffusion model to generate hand motion per primitive.

Key results

Limitations

Experimental setup

Datasets: OAKINK2

Benchmarks: OAKINK2

Metrics: MPJPE MPVPE PA-MPJPE AUC contact ratio solid intersection volume PSKL-J FID perceptual score

Baselines: METRO RLE KP-based Fit POEM

Builds on: OakInk MDM MoSH++ SMPL-X

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