In searching for possible solutions, Spark Lab decided to create a problem-specific tool. So this year, Stray Dogs direction was launched. It presupposes the first-ever use of neural networks to count stray animals in Yakutsk.
Timur Alexeev, Team Lead at Spark Lab, explained everything you might want to know about this novel system to control the stray dog population.
Why monitor dogs at all?
Besides preventing attacks from homeless canines, counting the population of stray animals is another important application. Using this data, you can identify the measures that can be taken to address the problem, "You have to know the extent of the problem to fight it effectively."
Another incentive is the fact that an evidence-based approach will help the city and the Republic justify additional funding for addressing the stray dog issue.
"Stray dog population monitoring is the foundation on which all further action stands. Understanding population size is what enables decision-making to plan and carry out interventions, as well as to evaluate their results. Our team has all the resources and capabilities to solve the task at hand. We plan to present the first results as early as late spring 2023," says the project manager Timur Alexeev.
How does it work?
Canine population monitoring will be done by processing data from the city's CCTV cameras. During the project implementation, we will use two different methods to estimate the animal population:
- Mathematical averaging, interpolation, and extrapolation over space and time;
- Machine learning methods to pull information from the city's infrastructure.
Computer vision will be the tool of choice for counting the canine population. Project team members are planning to teach the neural network to identify these four-legged animals directly from the video feed of outdoor CCTV cameras. The neural network itself will estimate the population of stray dogs and notify if a pack of dogs is detected somewhere. In other words, if the algorithm recognizes a big pack of dogs, the system will send the corresponding signal to the animal control services, so that they can respond quickly.
More than 20 people are already working on the project. Most of them are students from various Yakutsk universities, majoring in more than just tech and IT. Some participants are vets in training, providing invaluable insights into the behavior and habits of the animals. The project also features several experts in neural networking and computer vision who have experience in similar initiatives. They all share a common goal: to find a solution to this urgent problem and help safeguard the lives and health of the capital's residents.
In addition, the Spark Lab team plans to cooperate closely with all stakeholder agencies, including the city administration, the Department of Veterinary Medicine of the Republic of Sakha (Yakutia), the Republican Center of Infocommunication Technologies, etc.
Preparation for the project is underway, with the necessary work plan in place and agreements reached with R&D partners. Soon, the project will be trialed on a pilot site in Yakutsk, where the first tests of neural network algorithms will be carried out.
We wish good luck to the development team and will look forward to seeing the first results!