4 datasets found

A large-scale dataset for human-aware motion deblurring, consisting of 8,422 pairs of blurry and sharp images with 65,784 densely annotated FG human bounding boxes.

The commonly used benchmark for image deblurring is Go-Pro (Nah, Kim, and Lee 2017), where the input blurry images are synthetic.

The DVD dataset is a benchmark for image motion deblurring, containing 6708 synthetic blurry and sharp image pairs.

The GoPro dataset is used for training, which contains 2103 pairs of blurred clear control images. The validation set used for training is 460 pairs of randomly divided images....