MessyKitchens: Contact-rich object-level 3D scene reconstruction

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• The MessyKitchens dataset provides high-fidelity object-level ground truth in terms of 3D object shapes, poses and accurate object contacts.technical 95%
• The Multi-Object Decoder (MOD) approach demonstrates consistent and significant improvements over state-of-the-art methods on three datasets.result 90%
• MessyKitchens dataset significantly improves previous datasets in registration accuracy and inter-object penetration.result 90%
• The method extends the SAM 3D approach for single-object reconstruction with a Multi-Object Decoder for joint object-level scene reconstruction.method 95%
• The research addresses the need for physically-plausible scene reconstruction where objects obey physical principles of non-penetration and realistic contacts.technical 85%
• Reconstructing and decomposing common scenes into individual 3D objects remains a hard challenge due to large object variety, frequent occlusions and complex object relations.limitation 90%
• The method focuses on monocular 3D scene reconstruction from a single image.technical 95%
• The work advances object-level scene reconstruction along two directions: dataset creation and method development.method 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card