AmaraSpatial-10K: A Spatially and Semantically Aligned 3D Dataset for Spatial Computing and Embodied AI

TL;DR

AmaraSpatial-10K is a large, optimized 3D asset dataset designed for embodied AI and spatial computing, featuring evaluation tools and demonstrating significant improvements in retrieval and physics stability.

Contribution

The paper introduces AmaraSpatial-10K, a deployment-ready 3D asset dataset with evaluation metrics and demonstrates its advantages over existing datasets in retrieval and physics stability.

Findings

Improves CLIP Recall@5 by 3.4 times over Objaverse.

Achieves 99.1% physics-stability rate with Habitat-Sim.

Produces zero-overlap scenes as a drop-in asset bank.

Abstract

Web-scale 3D asset collections are abundant but rarely deployment-ready, suffering from arbitrary metric scaling, incorrect pivots, brittle geometry, and incomplete textures, defects that limit their use in embodied AI, robotics, and spatial computing. We present AmaraSpatial-10K, a dataset of over 10,000 synthetic 3D assets optimised for zero-shot deployment. Each asset ships as a metric-scaled, deterministically anchored .glb with separated PBR maps, a convex collision hull, a paired reference image, and multi-sentence text metadata. Alongside the dataset we introduce a reusable evaluation suite for 3D asset banks, a continuous Scale Plausibility Score (SPS), an LLM Concept Density metric, anchor-error auditing, and a cross-modal CLIP coherence protocol, and apply it to AmaraSpatial-10K alongside matched subsets of Objaverse, HSSD, ABO, and GSO. AmaraSpatial-10K improves CLIP Recall@5 by $3.4\times$ over Objaverse ($0.612$ vs. $0.181$, median rank $267 \rightarrow 3$), achieves a $99.1\%$ physics-stability rate under Habitat-Sim with $\sim 20\times$ wall-time speed-up, and produces zero-overlap scenes when used as a drop-in asset bank for Holodeck. Controlled ablations on the same asset bank attribute the retrieval gain to description richness.