We are pleased to invite you to a meeting of data analysis specialists, which will be held at the Avito Moscow office on October 28th. Mitap is dedicated to computer vision. You can learn about advanced achievements in the tasks of face recognition and image segmentation, about real-time adaptations of neural network and classical algorithms, and we will also present our service, AvitoNet. Detailed program and a link to the registration under the cat.

Schedule and speakers

AvitoNet: Avito computer vision service. Arthur Kuzin, Avito

Avito is actively using machine learning in various microservices and products. The report will tell about the part related to computer vision: AvitoNet. This is an internal dataset and neural networks trained on it. I will tell you about the tasks that were set before the AvitoNet service at the planning stage, as well as the chosen approach to the solution and the evolution of the service. Results will be shown with a demonstration of how AvitoNet works while submitting ads.

Image segmentation on the cutting edge of science. Evgeny Nizhibitsky, Rambler & Co

Recently, on popular sites, there are more and more contests related to computer vision, and especially with image segmentation. Some time ago, such approaches were dominated by U-Net based approaches everywhere. Eugene will tell about the decisions of two recently concluded competitions, where approaches based on the new network, claiming to be “popular recognition”, are well established.

Face recognition using deep neural networks. Sergey Milyaev, VisionLabs

The report will review the main scientific works that allowed neural networks to recognize people more accurately than people do. VisionLabs will also demonstrate the results of the independent international tests of the Face Recognition Vendor Test of the NIST Institute and the Labeled Faces in the Wild at the University of Massachusetts.

The use of computer vision for the analysis of sports events. Falaleev Nikolay, Constanta

A large amount of statistical information about the progress of sports games and competitions is collected manually. To automate this process, you can use computer vision techniques. The report will consider approaches to the analysis of the video stream from the game of table tennis in real time. Algorithms based on neural networks and classical computer vision will be proposed to determine the game events, the ball's flight path and record the results from the referee board, and the first results will be presented.

Passwords and attendances

Collection of participants and registration: 12:00
Reports start: 12:30
Address: Avito office, Lesnaya 7.

Registration is required .
UPD: Registration is closed, join the live broadcast of the event on the youtube channel AvitoTech .