CVPR2025
HELVIPAD: A Real-World Dataset for Omnidirectional Stereo Depth Estimation
Mehdi Zayene, Jannik Endres, Albias Havolli, Charles Corbière, Salim Cherkaoui, Alexandre Kontouli, Alexandre Alahi
Abstract
Figure 1. Examples from the HELVIPAD dataset in diverse settings (indoor, outdoor), under varied lighting (day, night). The dataset consists of paired top-bottom images with corresponding depth maps, captured from a robot navigating in dynamic human environments.