Jung-Ho Um, Juseop Kim, Hwan Suk Cheong, Journal of Internet Computing and Services, Vol. 25, No. 5, pp. 153-161, Oct. 2024
10.7472/jksii.2024.25.5.153, Full Text:
Keywords: agricultural data, Artificial intelligence, ONNX, interoperability, Data service

Abstract

With the development of artificial intelligence technology, various studies are being actively conducted to apply it to agriculture. In South Korea, AI training datasets are being produced for research on artificial intelligence technology. In various fields, including the agricultural and livestock sectors, datasets are posted through the site AI HUB. However, since AI models can be developed in multiple artificial intelligence learning frameworks, more things can interfere with the AI model as framework compatibility is considered in terms of serving the AI model. For system design, we derive requirements and propose the overall structure of the system. In addition, we show examples of the feasibility of implementing each component. To verify the feasibility of the proposed method, we trained a total of four eggplant disease classification models through transfer learning. We built a model with the best accuracy of 99% and converted it to ONNX, confirming that there was no difference in performance compared to the existing model. In future research, we plan to apply the model-sharing method verified in this paper to a data platform.

Statistics
Cite this article