Analysis of Modern Models Capabilities and Software for Automatic Annotation of Graphical Data in the Power Industry

The development of monitoring systems and the increase of graphic data amount leads to the fact that computer vision methods are actively used in various industries, including power industry. Computer vision is used to automatically analyze images and videos to detect and classify objects such as elements of power equipment. This considers the need to use computer vision methods in the power industry, areas of their use, describes related features and problems. The description of labeling methods, the most suitable architectures for detection, segmentation and classification tasks as well as an overview of modern software used for data labeling, is presented. Also, the paper contains the results of applying machine learning models with various augmentation tools for detection overhead power line insulators using an open dataset. Conclusions about the automatic method of labeling graphic data from the power industry are made. The obtained results demonstrate that the use of pre-trained models will only partially automate the labeling process.