Spring: L. Mehl, J. Schmalfuss, A. Jahedi, Y. Nalivayko, A. Bruhn — University of Stuttgart

RobustSpring: J. Schmalfuss, V. Oei, L. Mehl, M. Bartsch, S. Agnihotri, M. Keuper, A. Bruhn

Not logged in | Login | Create Account

Download Datasets

Our benchmark has two dataset components:

• Standard evaluation (required): Download the Standard Spring dataset below to measure accuracy on the clean test split.
• Robustness evaluation (optional): Download the Spring robustness dataset if you also want to evaluate performance under corruptions.

1. Standard Spring dataset
   Provides clean test-split ground truth for stereo, optical flow, and scene flow on 10 driving sequences.
   
   Download Standard Spring dataset
   (DOI: 10.18419/DARUS-3376)
2. Spring robustness dataset
   Contains the same sequences under 21 realistic corruptions (fog, noise, blur, etc.) for robustness evaluation.
   
   Download Spring robustness dataset
   (DOI: 10.18419/DARUS-5047)

After downloading, extract each archive into your chosen rootdir. Your folder tree will include entries like 0003/flow_FW_left/… for the standard split and <corruption>/0003/flow_FW_left/… for robustness. See the FAQ for full instructions on folder structure and submission.