FIELD: video analysis.
SUBSTANCE: method for preparing a set of videos for machine analysis that can be used in preparing videos and photographs intended for constructing articulated models of cattle animals. Video recording of animals as they pass along the corridor is carried out from 2 video cameras synchronously. The shooting is carried out from different angles: the body in profile on the right or left, the body 1/2 from the side, while each animal is photographed for at least 3 seconds. Videos are formed in good lighting with correct colour rendition, exposure is adjusted according to the animal’s body, shooting is carried out with shutter priority of no more than 1/125 s, and central single-point metering is used for exposure metering and focusing. The resulting videos are stored a in digital format in a machine-readable form on a storage medium.
EFFECT: ensuring creation of compact and effective datasets that enable accelerated training of neural networks, increasing the generalizing ability and stability of the functioning of neural networks, the ability to build articulated models of cattle animals, timely elimination of diseases, minimizing errors and simplifying the process of generating the required data sets for neural network algorithms.
1 cl
