This article implements an algorithm for recognizing the markings of a freight container based on deep neural networks. The main advantage of the proposed algorithm is the absence of the need for additional image transformations for marking localization and character segmentation. The EAST algorithm was used to solve the localization problem. To recognize the owner code and serial number of the container, Faster R-CNN Resnet 50 neural network models were trained. As a result of assessing the accuracy of the algorithm, appropriate conclusions were made, and possible options for improving the algorithm were formulated.
Keywords: machine learning, algorithm, cargo containers, neural network, image classification, recognition
This article describes training a neural network to recognize the digits of a freight container number. Due to the lack of a dataset containing a cargo container, training was performed on the Street View House Numbers dataset. Before training the model, the dataset was analyzed and histograms were built that reflect general information about the dataset. The neural network was trained in the Python programming language using the Tensorflow library. The obtained results of assessing the accuracy of the model operation allowed us to conclude that it is possible to use a data set and a neural network to solve the problem of recognizing the numbers of a freight container.
Keywords: machine learning, digit recognition, cargo container, neural network, image classification
`
Keywords: