This week you'll learn about Super-Resolution Generative Adversarial Networks (SRGAN).

Image

Visual arts are a very important part of humanity. It is used to express stories, emotions, and even feelings. When people talk about cult classic films like The Godfather from the 1970s, or even Ben Hur from the 1950s, they talk about the impact these films had on the cinema world. 

But there is a key point we fail to consider. These movies were made when technology hadn't yet come as far as it has today. Imagine trying to watch a Ben Hur print which probably has a resolution of (640×480) on a 4K Television (3840×2160). Old films aren't made for resolutions like these. You'll feel like Vito Corleone over here.

Using hardcoded algorithms might upscale the video, but the quality will become a big problem. Thankfully, like most things in the world, deep learning has come to the rescue.

In recent years, several ingenious deep learning algorithms have surfaced, which have shown tremendous improvement over classical algorithms, not only in terms of quality but also in terms of efficiency.  

This evolved a step further with the introduction of GANs in this domain. GANs, or Generative Adversarial Networks, are one of the biggest inventions in the deep learning world. And as expected, putting them to the task of upscaling low-resolution images to a higher resolution has yielded some spectacular results. 

The big picture: Today, we are combining the domain of super-resolution with GANs. 

How it works: The core idea of GANs is retained, that is, the adversarial loss is still being used in our task. But since super-resolution also deals with finer details located in the pixels of an image, an additional pixel loss is used to aid the final result. This is obtained by comparing the outputs of a VGG network.

Our thoughts: GANs add a phenomenal twist to the domain of super-resolution. It is extremely simple, as well as intuitive. 

Yes, but: The downside of GANs is their extreme need for data and computational power. SRGAN doesn't escape this, and thus we are forced to make additional arrangements to meet the requirements. 

Stay smart: Don't forget to try this out yourself! SRGAN is definitely not a concept to miss. 

Click here to read the full tutorial

Solve Your CV/DL problem this week (or weekend) with our Working Code

You can instantly access all of the code for Super-Resolution Generative Adversarial Networks (SRGAN) by joining PyImageSearch University. Get working code to

  1. Finish your project this weekend with our code
  2. Solve your thorniest coding problems at work this week to show off your expertise
  3. Publish groundbreaking research without multiple tries at coding the hard parts

Guaranteed Results: If you haven't accomplished your CV/DL goals, let us know within 30 days and get a full refund.

I want the code

Note: You may have missed this, but last Wednesday, we published a new post on Ensuring Your Research Stays Visible and General Tips.


The PyImageSearch Team


P.S. If you're interested in learning how to successfully apply deep learning to your own projects, I would recommend reading my book, Deep Learning for Computer Vision with Python.

Inside the book you'll find:
  • Super-practical walkthroughs that present solutions to actual real-world image classification (ResNet, VGG, etc.), object detection (Faster R-CNN, SSDs, RetinaNet, etc.), and segmentation (Mask R-CNN) problems
  • Hands on tutorials (with lots of code) that show you not only the algorithms behind deep learning for computer vision but their implementations as well.
  • A no-nonsense teaching style that is guaranteed to help you master deep learning for image understanding and visual recognition
If you're interested in learning more about the book, I'd be happy to send you a PDF containing the Table of Contents and a few sample chapters:

Click here to grab the PDF of sample chapters and Table of Contents

After clicking the above link, you'll receive a separate email with the PDF in a few short moments.