AIAny
Icon for item

Scanned Objects GLB

Converts 1,033 Google Scanned Objects household scans into web-ready GLB files, with KTX2 textures, LODs, and collision hulls for 3D scenes and simulation.

Introduction

Photorealistic objects are often the missing link between a simulated scene and the clutter of a real home. This collection makes more than a thousand scanned household products practical to load in web-based 3D viewers and simulation workflows: it preserves the original triangle count while shrinking the pack from 13.7 GB unpacked to about 344.8 MB.

What Sets It Apart
  • The models retain real product shapes and surface appearance rather than idealized CAD geometry, which can make synthetic scenes more visually representative.
  • Multiple delivery formats address different runtime constraints: WebP GLBs for straightforward viewing, KTX2 variants for compressed GPU textures, and three progressively simplified LODs for distance-based rendering.
  • Per-object metadata and convex collision hulls provide a useful starting point for scene assembly and physics, while the compact files lower the cost of distributing large object libraries.
Great Fit If...

Use it when you need varied, recognizable household objects for browser-based scenes, synthetic perception data, or robotics simulation. Look elsewhere if your task requires articulated parts, accurate measured mass, or physically faithful glass and reflective materials: these are static single-mesh scans, hulls are convex, and mass values are estimates.

Information

  • Websitehuggingface.co
  • Organizationscodepawl, Google Research
  • Published date2026/11/04

Categories

More Items

Turns a robot calibration episode and data request into synchronized, action-labelled training data: an agent builds the simulated task, while a video model renders realistic observations. Across five robot tasks, farm-trained policies scored 87 successes under unseen appearances, versus 46 for teleoperation-trained policies.

Evaluates AI agents on 56 research-based physics tasks, from literature review and experimentation to manuscript writing. It combines deterministic checks with expert rubrics to assess both correctness and research quality.

Provides minute-scale continuous gameplay video, camera trajectories, and structured annotations for training interactive world models that remain coherent over longer rollouts. Its two-minute sequences address the shortage of long-horizon training data.