NYUDv-2

The NYUDv-2 dataset is a benchmark that contains images with corresponding depth and semantic annotations, used to evaluate the performance of models in dense prediction tasks for spatial understanding.

Papers using NYUDv-2 (2)

M2H-MX: Multi-Task Dense Visual Perception for Real-Time Monocular Spatial Understanding2026V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning2026