|

Evaluation of the quality of object recognition on thermal imaging images using neural networks

Authors: Serbiev R.A., Berezan D.G.
Published in issue: #4(81)/2023
DOI: 10.18698/2541-8009-2023-4-881


Category: Informatics, Computer Engineering and Control | Chapter: System Analysis, Control, and Information Processing, Statistics

Keywords: Image classification, thermal imaging, object detection in images, image recognition, convolutional neural networks of deep learning, object search, GoogLeNet, Mask R-CNN
Published: 06.05.2023

A method for evaluating the quality of neural network recognition is presented. The full description of the Google Net neural network is given, and the principle of its operation is presented. A method of image recognition based on the classifying convolutional neural network Google Net, using the Viola-Jones method, is considered. The description of the Mask R-CNN neural network is given. Recognition of objects on thermal imaging images was carried out using the Mask R-CNN neural network. Experimentally, the results of a study of the accuracy of recognition of neural networks, depending on the learning parameters, were obtained. A comparative analysis of the results of modeling a neural network algorithm for solving the problem of classifying and searching for objects on thermal images is carried out.