The Helvipad dataset is a real-world stereo dataset designed for omnidirectional depth estimation. It comprises 39,553 paired equirectangular images captured using a top-bottom 360° camera setup and corresponding pixel-wise depth and disparity labels derived from LiDAR point clouds. The dataset spans diverse indoor and outdoor scenes under varying lighting conditions, including night-time environments.… See the full description on the dataset page:
https://huggingface.co/datasets/chcorbi/helvipad.
