1
$\begingroup$

When we have small low resolution, fuzzy image for example: enter image description here and if it to zoom, an unrelated set of pixels is obtained. For example enter image description here Tell me, please is there the way to train a neural network to recognize a magnified image. Clear it of pixels, so that it looks like this: (it is initial picture)

enter image description here

I want to understand:

  • What type of neural network is used for this?
  • What type data are needed to train neural network?
  • What should be the format of data?
$\endgroup$
  • 4
    $\begingroup$ i wanted to point out that all images are made of pixels -- you can't clear an image of pixels because then you'd be left with no image. $\endgroup$ – shimao Apr 10 at 21:41
  • 1
    $\begingroup$ Note that there are many algorithms for super-resolution, neural networks are not the only way. $\endgroup$ – Jan Kukacka Apr 11 at 8:46
4
$\begingroup$

It seems that you're asking about neural networks for single-image super-resolution. I think that your questions will be answered in a review paper on the topic, such as this one: Wenming Yang, Xuechen Zhang, Yapeng Tian, Wei Wang, Jing-Hao Xue, Qingmin Liao "Deep Learning for Single Image Super-Resolution: A Brief Review".

Single image super-resolution (SISR) is a notoriously challenging ill-posed problem, which aims to obtain a high-resolution (HR) output from one of its low-resolution (LR) versions. To solve the SISR problem, recently powerful deep learning algorithms have been employed and achieved the state-of-the-art performance. In this survey, we review representative deep learning-based SISR methods, and group them into two categories according to their major contributions to two essential aspects of SISR: the exploration of efficient neural network architectures for SISR, and the development of effective optimization objectives for deep SISR learning. For each category, a baseline is firstly established and several critical limitations of the baseline are summarized. Then representative works on overcoming these limitations are presented based on their original contents as well as our critical understandings and analyses, and relevant comparisons are conducted from a variety of perspectives. Finally we conclude this review with some vital current challenges and future trends in SISR leveraging deep learning algorithms.

To answer your specific questions:

What type of neural network is used for this?

Variants of convolutional neural networks, which stats.SE tags as . Particular approaches and networks are compared and contrasted in the article.

What type data are needed to train neural network?

According to the article, the typical procedure is to take a high-resolution image, blur/downsample the image to make a low-resolution image. The network's task is to attempt to reconstruct the high-resolution image from the low-resolution image.

What should be the format of data?

This question is too ill-posed to be answerable. You should use the format that's best for your purpose.

$\endgroup$
  • $\begingroup$ thank you. Off-top. Do you know where i can find dataset as example to try learn CNN? $\endgroup$ – Julia Apr 11 at 9:46
  • $\begingroup$ Modern neural networks software such as Tensorflow, Keras and Pytorch usually have a CNN demonstration script that works using the MNIST handwritten digits dataset. It's small and cheap enough that you can build a nice CNN on the data using a standard laptop computer. $\endgroup$ – Sycorax Apr 11 at 14:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.