Reproduce “ShAPO: Implicit Representations for Multi-Object Shape, Appearance, and Pose Optimization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Muhammad Zubair Irshad, Sergey Zakharov, Rares Ambrus, Thomas Kollar, Zsolt Kira, Adrien Gaidon
ShAPO is a unified framework that reconstructs 3D shapes and appearances and estimates 6D poses of multiple objects from a single RGB-D image. The system achieves improved performance on 6D pose estimation and 3D shape reconstruction by jointly optimizing these properties using implicit neural representations.
Key insights
- Jointly optimizing 3D shape, appearance, and 6D pose from a single RGB-D image leads to more coherent and accurate multi-object scene understanding.
- Disentangled latent representations for shape and appearance, combined with contrastive learning, enhance generalization to novel object instances.
- An octree-based differentiable optimization scheme enables efficient and progressive refinement of implicit surfaces, making multi-object reconstruction practical.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.