PBC:Image scaling/How it works: Difference between revisions

Latest revision as of 11:26, 30 January 2021

Comparison of scaling methods
Original photo	Upscaled 2x	Upscaled 4x	Upscaled 6x	Algorithm and description
600 × 863 reference	1,200 × 1,726 (2x)	2,400 × 3,452 (4x)	3,600 × 5,178 (6x)	Deep convolutional neural networks using perceptual loss. Developed on the basis of the super-resolution generative adversarial network (SRGAN) method, enhanced SRGAN (ESRGAN) is an incremental tweaking of the same generative adversarial network basis. Both methods rely on a perceptual loss function to evaluate training iterations.
304 × 443 reference	608 × 886 (2x)	1,216 × 1,772 (4x)	1,824 × 2,658 (6x)	Deep convolutional neural networks. Using machine learning, convincing details are generated as best guesses by learning common patterns from a training data set. The upscaled result is sometimes described as a hallucination because the information introduced may not correspond to the content of the source. Enhanced deep residual network (EDSR) methods have been developed by optimizing conventional residual neural network architecture. Programs that use this method include waifu2x, Imglarger and Neural Enhance.

@@ Line 11: / Line 11: @@
 |[[File:Markus Kage Paul Henry Serres Photography Upscale 4x.jpg|thumb|2,400 × 3,452 (4x)]]
 |[[File:Markus Kage Paul Henry Serres Photography Upscale 6x.jpg|thumb|3,600 × 5,178 (6x)]]
+|Deep convolutional neural networks using perceptual loss. Developed on the basis of the super-resolution generative adversarial network (SRGAN) method, enhanced SRGAN (ESRGAN) is an incremental tweaking of the same generative adversarial network basis. Both methods rely on a perceptual loss function to evaluate training iterations.
-|Deep convolutional neural networks using perceptual loss
 |-
 |[[File:Chad Driver Face Original.jpg|thumb|304 × 443 reference]]
@@ Line 17: / Line 17: @@
 |[[File:Chad Driver Face 4x.jpg|thumb|1,216 × 1,772 (4x)]]
 |[[File:Chad Driver Face 6x.jpg|thumb|1,824 × 2,658 (6x)]]
+|Deep convolutional neural networks. Using machine learning, convincing details are generated as best guesses by learning common patterns from a training data set. The upscaled result is sometimes described as a hallucination because the information introduced may not correspond to the content of the source. Enhanced deep residual network (EDSR) methods have been developed by optimizing conventional residual neural network architecture. Programs that use this method include waifu2x, Imglarger and Neural Enhance.
-|Deep convolutional neural networks
 |}