Generating an image training dataset for edges detection on an image using neural networks

One of the cornerstones of machine learning is training data. In case of the insufficient data, correct generalization becomes difficult. This problem is particularly serious when artificial neural networks are used. This work is devoted to an algorithm for generating parameters of arbitrarily shaped edges and creating images containing these edges. Part of the calculations can be performed at the GPU. It accelerates the overall generation more than 3 times. The dataset created in this way can be used in training artificial neural networks for the edge detection task.