Oleg Cherniy

CTO at RIA.com
About the Speaker:

Oleg holds a Magister Degree in Engineering and Design from Vinnytsia National Technical University. He is a co-founder and CTO at RIA.com. Oleg successfully writes his blog at Habr.

Current projects on which Oleg works:

Among numerous interesting projects, one of Oleg's favorites is Nomeroff Net - an automatic license plate recognition system. Nomeroff Net is a opensource python license plate recognition framework based on the application of a convolutional neural network on the Mask RCNN architecture, and customized OCR-module powered by GRU architecture.


Oleg is one of the authors and developers of the system. Oleg shares that until recently he was outraged by the current state of things on the market, because the technologies used in the process of recognition of license plates had been well-known for a long time, and what is more the market was commercial only, the quality and price for such services remained questionable though.

At AUTO.RIA.com, Oleg's team analyzes about 200,000 photos per day, so the decision was taken to fix the situation and develop the necessary module on their own. As far as Oleg is a supporter of open source systems, the code and data have been disclosed under open-label access under the GNU GPL.

About the report at the HighLoad UA conference:

How to increase the value of your data using machine learning.

The example can be seen in the case of AUTO.RIA. The algorithm for finding license plates in the photo analyzes up to 200 000 photos per day.

Oleg will talk about data that lacks being payed attention to during the analysis in traditional ways. There will be presented several ideas of the use of machine learning techniques that will help improve the interpretation of the data that users are sending. Oleg will comment the problems of using modern ML-models in the mode of the inference, when it is necessary to find the balance between the hardware resources and the speed of information processing on the example of practical cases of the site AUTO.RIA.com

Follow the HighLoad UA news conference on the Facebook page
Oleg Cherniy
CTO at RIA.com

About the Speaker:

Oleg holds a Magister Degree in Engineering and Design from Vinnytsia National Technical University. He is a co-founder and CTO at RIA.com. Oleg successfully writes his blog at Habr.
Current projects on which Oleg works:

Among numerous interesting projects, one of Oleg's favorites is Nomeroff Net - an automatic license plate recognition system. Nomeroff Net is a opensource python license plate recognition framework based on the application of a convolutional neural network on the Mask RCNN architecture, and customized OCR-module powered by GRU architecture.

Oleg is one of the authors and developers of the system. Oleg shares that until recently he was outraged by the current state of things on the market, because the technologies used in the process of recognition of license plates had been well-known for a long time, and what is more the market was commercial only, the quality and price for such services remained questionable though.

At AUTO.RIA.com, Oleg's team analyzes about 200,000 photos per day, so the decision was taken to fix the situation and develop the necessary module on their own. As far as Oleg is a supporter of open source systems, the code and data have been disclosed under open-label access under the GNU GPL.
About the report at the HighLoad UA conference:
How to increase the value of your data using machine learning.

The example can be seen in the case of AUTO.RIA. The algorithm for finding license plates in the photo analyzes up to 200 000 photos per day.

Oleg will talk about data that lacks being payed attention to during the analysis in traditional ways. There will be presented several ideas of the use of machine learning techniques that will help improve the interpretation of the data that users are sending. Oleg will comment the problems of using modern ML-models in the mode of the inference, when it is necessary to find the balance between the hardware resources and the speed of information processing on the example of practical cases of the site AUTO.RIA.com
Follow the HighLoad UA news conference on the Facebook page
Close
Close
Зв'язок із нами
Telegram
Messenger
Mail
Реєстрація