The UFO-120 Dataset

Dataset for Simultaneous Enhancement and Super-Resolution (SESR) of underwater imagery.

Introduced in the following paper:  https://arxiv.org/pdf/2002.01155.pdf

Folder structure:

  • train_val/ contains 1500 paired samples for training/validation
  • TEST/ contains 120 paired samples for benchmark evaluation

Paired data for each set:    

  • hr/ contains high-resolution ground truth images: Y
  • mask/ contains ground truth saliency maps: S
  • lrd/ contains low-resolution distorted images: X  

Possible Usage:    

  • Train SESR models: {X} → {Y}
  • Train image enhancement models: {X} → {Y} (downscaled)
  • Train image super-resolution models: {Y} (downscaled) → {Y}
  • Train saliency prediction models: {Y} (downscaled) → {S}

The Y images are of size 640 x 480. The X and S are of size 320 x 240. Hence, the default configuration is for training 2x models; rescale X and for other configurations: 3x, 4x, etc.