Creation Method of Priori Neural Network Data Set for Processing Digital Aerial Photographs in Automatic Mode

The availability of large data sets contributes to the rapid expansion of the deep learning methods in general and computer vision methods in particular. At the same time, there is a lack of training data in many areas, which becomes an obstacle to the practical application of computer vision models. The article proposes a creation method of the necessary set of a priori objects images obtained by aerial photography from unmanned aerial vehicles, which differs from existing ones by adaptation to the shooting factors and the specifics of thematic processing objects. The use of the proposed method will allow to significantly reduce the complexity of required data collecting and add increasing techniques, which require less computing resources and enhance the object detection reliability.