Software for large format aerospace image marking and training samples preparation

Автор: Gavrilov Dmitriy Aleksandrovitch, Schelkunov N.N.

Журнал: Научное приборостроение @nauchnoe-priborostroenie

Рубрика: Информатика, вычислительная техника и управление

Статья в выпуске: 2 т.30, 2020 года.

Бесплатный доступ

This paper presents software designed for marking objects on aerospace images in visible and infrared spectra, creating a base of samples, and preparing training samples to solve the problem of detecting and localizing objects. The main functionality of the program consists of searching objects of interest, viewing an image, measuring the size of objects, marking up objects of interest on images using special marking tools, editing previously made marking of objects, designation and marking objects. Precedent information for training is aerospace images with images of objects marked on them, highlighted with the greatest possible accuracy. The markup program works in automatic, semi-automatic and manual modes. Work in semi-automatic mode allows the operator to clarify the localization and marking of objects and greatly facilitates and improves the quality of preparation of training samples for the subsequent training of detection and classification algorithms.

Еще

Markup of training samples, neural network training, active training

Короткий адрес: https://sciup.org/142223748

IDR: 142223748   |   DOI: 10.18358/np-30-2-i6775

Статья научная