Image processing *

This article continues a series of notes about colorization. During today's experiment, we’ll be comparing a recent neural network with the good old Deoldify to gauge the rate at which the future is approaching.

This is a practical project, so we won’t pay extra attention to the underlying philosophy of the Transformer architecture. Besides, any attempt to explain the principles of its operation to a wide public in hand waving terms would become misguiding.

A lecturer: Mr. Petrov! How does a transformer work?
Petrov with a bass voice: Hum-m-m-m.

Google Colorizing Transformer vs Deoldify

With the advent of mobile phones with high-quality cameras, we started making more and more pictures and videos of bright and memorable moments in our lives. Many of us have photo archives that extend back over decades and comprise thousands of pictures which makes them increasingly difficult to navigate through. Just remember how long it took to find a picture of interest just a few years ago.

One of Mail.ru Cloud’s objectives is to provide the handiest means for accessing and searching your own photo and video archives. For this purpose, we at Mail.ru Computer Vision Team have created and implemented systems for smart image processing: search by object, by scene, by face, etc. Another spectacular technology is landmark recognition. Today, I am going to tell you how we made this a reality using Deep Learning.

Hi everybody! I’m a research engineer at the Mail.ru Group computer vision team. In this article, I’m going to tell a story of how we’ve created AI-based photo restoration project for old military photos. What is «photo restoration»? It consists of three steps:

• we find all the image defects: fractures, scuffs, holes;
  
• we inpaint the discovered defects, based on the pixel values around them;
  
• we colorize the image.
  

Further, I’ll describe every step of photo restoration and tell you how we got our data, what nets we trained, what we accomplished, and what mistakes we made.

With the recent progress in Neural Networks in general and image Recognition particularly, it might seem that creating an NN-based application for image recognition is a simple routine operation. Well, to some extent it is true: if you can imagine an application of image recognition, then most likely someone have already did something similar. All you need to do is to Google it up and to repeat.

However, there are still countless little details that… they are not insolvable, no. They simply take too much of your time, especially if you are a beginner. What would be of help is a step-by-step project, done right in front of you, start to end. A project that does not contain «this part is obvious so let's skip it» statements. Well, almost :)

In this tutorial we are going to walk through a Dog Breed Identifier: we will create and teach a Neural Network, then we will port it to Java for Android and publish on Google Play.

For those of you who want to see a end result, here is the link to NeuroDog App on Google Play.

Web site with my robotics: robotics.snowcron.com.
Web site with: NeuroDog User Guide.

Here is a screenshot of the program: