Application of the Canny’s operation and the template matching method to improve the training set in computer vision training’s issue system
Mikhail G. Dorrer, A E Tolmacheva · Journal of Physics Conference Series · 2020
Abstract This article focuses on solving the problem of increasing the efficiency of the process of forming a training set in computer vision training’s issue system based on an artificial neural network. The solution to this problem is implemented by developing software for searching for contours using the canny method in combination with the method of searching for identical samples and then deleting them by using the templatematching algorithm. The implementation of a software tool for expeditious preparation of a high-quality training sample for training a neural network based on the YOLOv3 architecture is described. As the initial data, we used a video stream received from cameras located in the working area of computer vision - on the shelf of a smart refrigerator of the Robimarket system.