HM-3D-v-2

The 'HM-3D-v-2' dataset/benchmark contains a collection of 3D environments designed to evaluate object navigation tasks, specifically focusing on how effectively robots can search for unobserved targets in unknown settings by mimicking human-like exploration strategies.

Papers using HM-3D-v-2 (2)

IntentNav: Learning Spatial-Visual Object Navigation from Human Demonstrations2026DRIVE-Nav: Directional Reasoning, Inspection, and Verification for Efficient Open-Vocabulary Navigation2026