INFORMATION SYSTEM FOR IMAGE CLASSIFICATION AND LABELING FOR TRAINING ARTIFICIAL INTELLIGENCE MODELS
DOI:
https://doi.org/10.35546/kntu2078-4481.2023.3.17Keywords:
image classification, image labeling, artificial neural networks, artificial intelligence model, information technology, automationAbstract
The paper is devoted to the development of an information technology for image classification and labeling for the purpose of training artificial intelligence models. Improving the speed and accuracy of image classification and labeling by assigning them specific tags or categories opens up new possibilities for the use of machine learning in various fields, including computer vision, medical diagnostics, and image recognition. The analysis of existing image annotation systems has shown that their weaknesses lie in the incompleteness and inconvenience of implemented tools, as well as insufficiently high execution speed. The proposed approach for image classification and labeling involves the use of artificial neural network technologies. For the automation of image classification, the ResNet network was selected, which is trained within the framework of a single dataset, thus reducing the time required for the operation. For image labeling tasks, the SAM network was applied, which allows for generalizing unfamiliar objects and images without the need for additional training. Research on the use of these technologies on a test dataset has demonstrated their sufficiently high accuracy. Requirements for an information system for automating image classification and labeling have been formulated, which are formalized in the form of a UML use case diagram. The system's structure has been designed, and development tools have been chosen. The software has been created using the Python programming language and subjected to testing. MongoDB has been selected as the database management system due to its free-ofcharge availability and productivity. The research results can be used by information technology developers working in the field of artificial intelligence model training.
References
Data Labeling: How to Choose a Data Labeling Partner in 2023 [Електронний ресурс]. Режим доступу до ресурсу: https://research.aimultiple.com/data-labeling/#why-is-it-important-now
Acs B, Rantalainen M, Hartman J. Artificial intelligence as the next step towards precision pathology. J Intern Med. 2020. Vol. 288. P. 62–81.
Gulbahar Karatas. Data Annotation in 2023: Why it matters & Top 8 Best Practices. AIMultiple [Електронний ресурс]. Режим доступу до ресурсу: https://research.aimultiple.com/data-annotation/.
Документація Computer Vision Annotation Tool (CVAT) [Електронний ресурс]. Режим доступу до ресурсу: https://github.com/opencv/cvat.
Документація Label studio(LS) [Електронний ресурс]. Режим доступу до ресурсу: https://labelstud.io/guide/.
Abdou, M.A. Literature review: efficient deep neural networks techniques for medical image analysis. Neural Comput & Applic. 2022. Vol. 34. P. 5791–5812. https://doi.org/10.1007/s00521-022-06960-9.
Gulzar, Y. Fruit Image Classification Model Based on MobileNetV2 with Deep Transfer Learning Technique. Sustainability. 2023. Vol. 15, 1906. https://doi.org/10.3390/su15031906.
Dhiman, G., Kumar, A.V., Nirmalan, R. et al. Multi-modal active learning with deep reinforcement learning for target feature extraction in multi-media image processing applications. Multimed Tools Appl Vol. 82. P. 5343–5367. https://doi.org/10.1007/s11042-022-12178-7.
G. Huang, Z. Liu, L. Van Der Maaten and K. Q. Weinberger. Densely Connected Convolutional Networks. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017, pp. 2261-2269. https://doi.org/10.1109/CVPR.2017.243.
Огляд ResNet та його варіантів [Електронний ресурс] – Режим доступу до ресурсу: https://datascience.eu/machine-learning/an-overview-of-resnet-and-its-variants/.
Офіційний Сайт Segment Anything Model (SAM) [Електронний ресурс]. Режим доступу до ресурсу: https://segment-anything.com/.
