Artificial intelligence will help to estimate the dog age from a photo

Artificial intelligence will help to estimate the dog age from a photo

ETU researchers will co-organize international competition on automatic recognition of the dog age in the framework of the European Conference on Artificial Intelligence.

26.04.2019 1872

The DogAge Challenge will take place as part of ICANN19, the largest conference on artificial intelligence in Europe, which is held by the European Neural Network Society for the 28th time in September 2019 in Munich (Germany).

Researchers of the St. Petersburg Electrotechnical University "LETI" together with colleagues from Tech4Animals Lab, University of Haifa, and School of Environmental and Life Sciences, University of Salford, will hold the competition called “The DogAge Challenge - Automatic Dog Age Estimation.” Its main task is to develop a neural network able to classify dogs by age groups based on their photos.

Young researchers of ETU “LETI” Dmitry Kaplun, Associate Professor of the Department of Automation and Control Processes, and Alexander Sinitsa, graduate student, initiated the competition, which will take place for the first time.

“ETU has long been developing in the field of artificial intelligence and machine learning. In particular, these tools are widely used in computer vision technology, which recognizes various objects and their properties based on visual data. This study is part of a large interdisciplinary project on the analysis of animal state and behavior based on visual data, which is being carried out jointly with colleagues from the University of Haifa (Israel) and the University of Salford (United Kingdom). Earlier at the HEALTHINF 2019 conference, dedicated to health informatics, the Automated System for Analyzing Animal Behavior, which assesses the quality of dog sleep, have been already successfully presented in Prague.”

Dmitry Kaplun, Associate Professor of the Department of Automation and Control Processes

Call for applications starts May 1, 2019. Anyone can participate in the competition. Both individual and collective work are accepted. One author or team can submit only one work.

Participants will have to assess the age of the dog from a photo and assign the data to one of three groups: young (0-2 years), adult (2-5 years), senior (over 6 years old) with an accuracy not lower than the one specified. Accuracy is determined using the metrics set by the organizers. Organizers will provide a dataset of dog photos collected by biologists from the University of Salford (Manchester, UK).

The quality of solutions will be measured using the mean average precision (mAP). The jury will include representatives of ETU "LETI," the University of Haifa, and the University of Salford.

Winners and prizewinners will receive cash prizes between 500 and 1000 euros.

Solutions are accepted from May 1 to September 10, 2019, by email Requirements for registration will soon appear on the website of the competition.

The winner will be announced at the ICANN19 conference in Munich (Germany) on September 17-19, 2019.

Details on the conference website