Spring Dataset and Benchmark Logo

L. Mehl, J. Schmalfuss, A. Jahedi, Y. Nalivayko, A. Bruhn

Not logged in | Login | Create Account

Welcome to the Spring dataset and evaluation benchmark for stereo, optical flow and scene flow estimation!

Download dataset







Download sample






Teaser image
The Spring dataset consists of high-resolution left and right stereo images (1920x1080px). It also contains full scene flow data with 4x super-resolution (3840x2160px). For stereo/depth estimation, left-to-right and right-to-left disparity is given for every frame (see second row). For scene flow estimation, the dataset provides disparity change for left and right as well as forward and backward direction (see third row). For optical flow estimation, the Spring dataset contains left and right, forward and backward optical flow (see last row).



If you make use of our dataset or benchmark results, please cite our paper:

    author    = {Lukas Mehl and Jenny Schmalfuss and Azin Jahedi and Yaroslava Nalivayko and Andr\'es Bruhn},
    title     = {Spring: A High-Resolution High-Detail Dataset and Benchmark for Scene Flow, Optical Flow and Stereo},
    booktitle = {Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    year      = {2023}

Further benchmarks

There are many benchmarks that have been pushing forward research in the domains of motion estimation and stereo. Most notable examples are the Middlebury optical flow and stereo benchmark, the KITTI 2012 optical flow and stereo benchmark, the Sintel optical flow benchmark, KITTI 2015 as the first benchmark for scene flow, optical flow and stereo, the ETH3D stereo benchmark and the VIPER optical flow benchmark. As a great addition to existing benchmarks, the Robust Vision Challenge ranks algorithms according to their cross-benchmark generalization. Unfortunately, the HD1K benchmark seems to be offline at the time of writing.