Reproduce “Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items” 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
Laura Downs, Anthony Francis, Nate Koenig, Brandon Kinman, Ryan Hickman, Krista Reymann, Thomas B. McHugh, Vincent Vanhoucke
The Google Scanned Objects (GSO) dataset provides over 1,000 high-quality 3D scanned household items, released under an open-source license, to enhance realism in robotics simulation. This collection supports research in object recognition, grasping, and synthetic data generation by bridging the gap between virtual training and real-world deployment.
Key insights
- A tailored scanning pipeline, integrating structured light, HDR imaging, and robust mesh processing, can efficiently produce a large volume of high-fidelity 3D object scans.
- Meticulous curation, including calculating physical properties and generating simulation-ready formats, is crucial for making 3D datasets immediately usable in robotics simulators.
- Providing over 1,000 diverse, high-quality, and open-licensed 3D household objects addresses a critical need for realistic virtual environments, enabling better sim-to-real transfer for robot learning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.