Nah et al. 2017 dataset

The dataset used in the paper is a synthetic dataset of blurred videos, generated by capturing sharp videos at high frame rate and averaging adjacent frames.

BibTex: