360° Stereo Video Test Datasets

360-degree cameras are revolutionising a multitude of applications and providing a huge boost for the development of others related to Virtual/Augmented/Mixed Reality. Virtual Reality (VR) is aimed to provide immersive experiences, where higher realism of captured or reproduced visual scenes and means for natural interaction with 3D content are essential.

However, there are still several limitations and issues necessary to address in this area such as distortions, panoramas stitching, depth estimation, and degrees of freedom provided to the final user.

The datasets presented by CIVIT provide 4 or 8 different fisheye views generated from a Nokia OZO. This allows testing of multiple algorithms like depth estimation or panoramas stitching. 

All the available dataset are composed of fisheye views, stitched videos, and metadata. Please, find the citation instructions on the dataset page

Dataset: Climbing

Number of Fisheye views: 4
Length of the video: 01:47 minutes
Resolution: 2K x 2K

Dataset: Pole Vault

Number of Fisheye views: 8
Length of the video: 00:38 minutes
Resolution: 2K x 2K 

Dataset: Bicycle

Number of Fisheye views: 8
Length of the video: 00:54 minutes
Resolution: 2K x 2K 

Dataset: Circle

Number of Fisheye views: 8
Length of the video: 02:23 minutes
Resolution: 2K x 2K

Dataset: Clapping song

Number of Fisheye views: 4
Length of the video: 01:22 minutes
Resolution: 2K x 2K

These datasets are licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. For commercial use please contact civit@tuni.fi