Workshop paper
4–6 main pages submitted for peer review and presentation at the NeurIPS workshop.
RoCo-Spring: The Robust Correspondence Challenge
RoCo-Spring 2026 is a NeurIPS 2026 challenge focused on evaluating robust dense correspondence methods under realistic distribution shifts, targeting Optical Flow, Stereo Matching, and Scene Flow. It introduces the RobustSpring dataset with 20 real-world corruptions to assess model reliability in safety-critical applications like autonomous driving and robotics, alongside the high-resolution Spring benchmark for clean performance.
Paper fit
A strong submission should clearly identify its contribution and evaluate it appropriately.
4–6 main pages submitted for peer review and presentation at the NeurIPS workshop.
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026