license: cc-by-4.0
tags:
- super-resolution
pretty_name: DF2K_BHI
size_categories:
- 1K<n<10K
DF2K_BHI SISR Dataset
This is a filtered version of the DF2K dataset, specifically aimed for single image super-resolution model training.
It is a result, or is one of the testing sets, for my huggingface community blog post on my BHI-Filtering method.
[TODO CHECK IF THIS IS IC>=0.4 or IC>= 0.5; it is 6'620 tiles currently]
Dataset Details
This is based on the DF2K dataset, which consists of X images.
These images then have been tiled into X 512x512px tiles for faster sisr I/O training speed.
The BHI filtering has been applied with the following default threshold values:
Blockiness < 30
HyperIQA >= 0.2
IC9600 >= 0.4
Which results in this DF2K_BHI dataset with X 512x512 training tiles.
Visualization
Here is a visual example of the first few tiles that are in the dataset:
[TODO PUT AN IMAGE HERE THAT SHOWS SOME TILES OF THE DATASET]
Training
Here is the training validation metrics on the DIV2K validation dataset, using the plksr_tiny architecture option, in comparison to the tiled full DF2K dataset:
Tiled DF2K dataset (here called X): X tiles DF2K_BHI dataset: X tiles
[TODO PUT IMAGES HERE OF TENSORBOARD METRICS OF ONLY TILED DF2K AND DF2K_BHI]