Computer vision and machine learning in PHP using the library opencv

Computer vision and machine learning in PHP using the library opencvHello. This is my jubilee article on the habr. For almost 7 years I wrote 10 articles (including this one), 8 of them - technical. The total number of views of all articles is about half a million.
 
I made the main contribution to two hubs: PHP and Server Administration. I like to work at the junction of these two areas, but the scope of my interests is much broader.
 
Like many developers, I often use the results of someone else's work (articles on the habr, code on the githaba, ), so I'm always glad to share with the community my results in response. Writing articles is not only a return of ...
+ 0 -

Organization of the university course on signal processing

Organization of the university course on signal processingPedagogy interested me a very long time and, for many years, I, being a student, brought up, but at the same time harassed and jerked by the existing training organization, thought about how to improve it. Recently, I have been given an opportunity to check out some of the ideas in practice. In particular, this spring I was given an opportunity to read the course "Signal Processing" at the Polytechnic University (SPbPU). His organization, especially the organization of reporting - the first experiment, the results of which seem to me to be somehow successful, and in this article I want to tell ...
+ 0 -

One day from the life of the Eagle in a few seconds

One day from the life of the Eagle in a few secondsCurrently, on the Internet there are many sites where you can watch videos online from CCTV cameras, which are located in public places of certain cities. In particular, several such cameras are installed in the city of Orel. One of the cameras, which is located on one of the high-rise buildings per. Postal, aimed at the Children's Park, appeared the very first. At least on one of the sites in the spring in 2011 I found a page where you could watch the video from this camera "live" (online). However, this broadcast was organized very interesting. Video came with a very low frame rate - 2 frames ...
+ 0 -

Kaggle: Amazon from Space - tricks and khaki for learning neural networks

Kaggle: Amazon from Space - tricks and khaki for learning neural networks  
 
Last summer, was over. competition on the kaggle site, which was devoted to the classification of satellite imagery of the Amazon forests. Our team took 7th place out of 900+ participants. Despite the fact that the competition ended a long time ago, almost all the methods of our decision are applicable so far, not only for competitions, but also for training neural networks for selling. For details under cat.
 
albu [/i]
 
At the time of participation in the competition, he directed all ml of the direction in the company Geoscan. But since then ...
+ 0 -

"Eye in the sky": patrol drones with the recognition of violence in clusters of people and public places

"Eye in the sky": patrol drones with the recognition of violence in clusters of people and public places  

The illustration on the left shows 14 key points on the human body that recognize the machine vision system: head, neck, shoulders, elbows, wrists, hips, knees, ankles. On the right, above the drone Parrot AR with the system of violence detection. On the bottom right are the individual photo elements from the training dataset with key points

 
 
In our time, UAVs are increasingly being used by law enforcement agencies and special services. Usually for espionage, reconnaissance, border control, etc. The police are not yet actively using drones to patrol city streets. But there is a huge potential here...
+ 0 -

Humansnotinvited: solve the catch on bash

Humansnotinvited: solve the catch on bashGreetings, dear reader!
 
 
Many of you have met with the CAPTCHA - the automatic Turing test. It allows you to separate real people from different bots. Recently, reCAPTCHA from Google Inc. has become very popular. On it you must select images containing certain objects, for example, cars. Relatively recently appeared site It does exactly the opposite: it separates the bots from people.
 
Reddit still managed to pass the captcha.
 
 

Let's start learning


 
If you pay attention, the CAPTCHA has a very limited number of topics on the site (what you need to find in the picture)...[/h]
+ 0 -

Recognition of scenes on images with the help of deep convolutional neural networks

Many products of our company work with images. Some time ago we decided to add to such services an "intelligent" search for photos, their tagging. Such functionality will be included in the Computer Vision API for further use in the company's products. One of the important ways of tagging images is tagging by scenes, when we get something like this:
 
 
Recognition of scenes on images with the help of deep convolutional neural networks  
ImageNet , and algorithms (convolutional neural networks from 2012), which won in the discipline of object recognition, as a rule, became state-of-the-art. However, the recognition of scenes has been studied much less, only since last year this task ...
+ 0 -

Improving the quality of images using a neural network

Today, I want to talk about an interesting approach to improve image quality. The official name of the Super Resolution approach.
 
Improvement of image quality by software methods is known from the beginning of the appearance of digital images, but in the last 3 years there has been a qualitative jump caused by the use of neural networks.
 
 
Improving the quality of images using a neural network  
 
An example of improving image quality using Super Resolution technology.
 
arxiv.org/pdf/???v3.pdf
 
Image Super-Resolution Using Deep Convolutional Networks
 
[2] arxiv.org/pdf/???.pdf
 
Image Restoration ...
+ 0 -

Image processing: Tensorflow Object Detection API

Image processing: Tensorflow Object Detection APIThe last few years in the development of deep neural networks there is a real revolution: new architectures are emerging, frameworks for developers are being improved, and iron for experiments can be obtained completely free of charge - for example, within the framework of project Google colaboratory . To everyone who is interested in how to apply pre-baked models from the repository Tensorflow Object Detection API to the decision of the task, using capacities Colaboratory - welcome under kat.
 
card. Tesla K80 .
 
 
You ...
+ 0 -

USB3Vision and GenICam. A look from the inside. I

USB3Vision and GenICam. A look from the inside. I

Introduction


 
The modern world is difficult to imagine without video cameras. They are so densely settled in our lives that they have become an integral part of it, whether we like it or not. Smartphones, computers, security systems, etc. The list of spheres of application can be continued for a long time, but in the final analysis the goal is pursued one - the construction of an image based on the light information coming from the surrounding world to the photosensitive sensor.
 
 
Ubiquitous industrialization, optimization of production capacities, development of production equipment for ...[/h]
+ 0 -