SYNTHESIZED DATA CREATION SOFTWARE AND FEEDBACK SIMULATOR FOR TESTING MACHINE LEARNING ALGORITHMS
Автор: A. G. Lapushkin, D. A. Gavrilov, O. A. Potkin
Журнал: Научное приборостроение @nauchnoe-priborostroenie
Рубрика: Математические методы и моделирование в приборостроении
Статья в выпуске: 1, 2023 года.
Бесплатный доступ
This paper presents software that allows the creation of new training data and tests existing algorithms in simulation mode. The developed program allows obtaining related data sets, including combined sets of the visible and infrared ranges using a single camera or stereo pair, additional information in the form of lidar data or a depth map, a segmentation pattern, and data on the location of objects of interest in a photo or video image. The structure of the developed software allows for further improvement of the approaches and possibilities for finalizing the resulting pipeline for different purposes and tasks.
Machine learning, training sets, training of neural network algorithms, algorithm testing, simulator
Короткий адрес: https://sciup.org/142236955
IDR: 142236955