Every year, hundreds of millions of juvenile omuls are released in Baikal and the rivers flowing into it. One of the companies specializing in the cultivation and release of this fish species larvae into the natural environment is the Bolsherechensky Fish Hatchery. The Hatchery specialists catch spawning fish for efficient, centralized caviar collection. The eggs are placed in incubators, where they develop into larvae without external threats. At the end of spring there is a mass egg hatching. Before releasing them into the natural environment, specialists need to count them.
Currently, the omul larvae at the Bolsherechensky Hatchery are counted using the volumetric method on the Weiss apparatus. This is a labor-intensive process that has been used for at least 40 years. There is also equipment for automatic counting, but it is not operational and is calculated for shrimp or medium-sized fish rather than for such small objects as omul larvae that have a linear size up to 10 mm. An alternative method will be proposed by NSU specialists.
Pavel Zhikharev, Senior Programmer at the NSU College of Informatics, explained,
The problem with the old counting device is that it was created to search for objects of a certain size and shape, that is, only spots or silhouettes are visible. Therefore, it may count a small stone larva or an air bubble, but the larva itself can be missed. We will train neural networks to recognize the omul larva in the same way as the human eye.
During a visit to the Hatchery, University representatives became acquainted with the current technological procedure for the incubation of fish caviar in order to take into account the specifics of the fishery when they develop equipment to count the larvae. During the trip, it became clear to the specialists that they also needed to develop a system for distributing the flow of water and larvae in order to correctly count the number of small objects moving in a continuous mass.
Zhikharev added,
A similar neural network accounting system developed at 吃瓜官网, is already being tested for counting glass containers on a conveyor. However, in the case of fish farming, a number of additional industry-specific factors need to be taken into account. With these features in mind, we already tested the device prototype for counting larvae on several test videos and the accuracy was 98%.
Experts predict that they will have time to complete work on the development and installation of equipment before the omul larvae hatching season next spring.
This work is being conducted under the Priority 2030 program.
