Comparison of Deep Learning Approaches for Detecting Diseased Potato Plants

The timely identification of diseased agricultural crops is essential for maintaining food security and reducing economic losses. The integration of machine vision with deep learning algorithms offers a more efficient and accurate method for monitoring potato crops and detecting disease symptoms than conventional visual assessment techniques. (Research purpose) This study aims to conduct a comparative analysis of one-stage and two-stage deep learning approaches for recognizing diseased and healthy potato plants. (Materials and methods) Two approaches were employed to train neural networks for the identification of diseased and healthy potato plants: a one-stage and a two-stage approach. In the one-stage approach, a single deep learning algorithm was used to simultaneously perform plant classification and localization. The two-stage approach utilized two separate algorithms: the first was responsible for detecting plant boundaries, while the second classified the identified regions as healthy or diseased. The models were trained on diverse datasets comprising images of individual potato leaves as well as entire plants. (Results and discussion) A comparative analysis was performed to evaluate the effectiveness of the one-stage and two-stage deep learning approaches in detecting diseased potato plants. For each training method, both the overall mean squared error and the coordinate-specific mean squared error were computed. Additionally, confusion matrices were generated to assess classification performance. The analysis revealed differences in accuracy and precision between the two approaches, highlighting their respective strengths and limitations. (Conclusions) The two-stage approach proved to be highly effective in distinguishing between diseased and healthy potato plants. Although it exhibited a slight reduction in coordinate-prediction accuracy – particularly when trained on both individual leaf images and whole-plant images – it offered superior classification performance. Both approaches demonstrate distinct advantages and hold significant potential for integration with modern technologies aimed at enhancing the early detection of phytopathologies in agricultural crops.

1. Borychev S.N., Vladimirov A.F., Koloshein D.V. et al. To the question of research on storing potatoes. Bulletin of the Ryazan State Agrotechnological University named after P.A. Kostychev. 2019. N2(42). 129-134 (In Russian). EDN: HIKKNU.

2. Tarkhanova Z.E. Food security of the state: content, significance, threats, food security. Economics and Management: Problems, Solutions. 2024. Vol. 6. N10(151). 84-90 (In Russian). DOI: 10.36871/ek.up.p.r.2024.10.06.010.

3. Izmaylov A.Yu., Lobachevskiy Ya.P., Dorokhov A.S. et al. Modern agriculture technologies and equipment – trends of an agritechnika 2019 exhibition. Tractors and Agricultural Machinery. 2020. N6. 28-40 (In Russian). DOI: 10.31992/0321-4443-2020-6-28-40.

4. Dorokhov A.S., Sibirev A.V., Ponomarev A.G., Sazonov N.V. Analytical justification of the technological process of the machine for removing infected potato and vegetable crops. Agrarian Scientific Journal. 2024. N5. 130-136 (In Russian). DOI: 10.28983/asj.y2024i5pp130-136.

5. Lobachevsky Ya.P., Tsench Yu.S. Principles of forming machine and technology systems for integrated mechanization and automation of technological processes in crop production. Agricultural Machinery and Technologies. 2022. Vol. 16. N4. 4-12 (In Russian). DOI: 10.22314/2073-7599-2022-16-4-4-12.

6. Alferyev D.A. Practic of implementing convolutional neural networks in agriculture and agro-industrial complex. Agricultural and Livestock Technology. 2020. Vol. 3. N2. 4 (In Russian). DOI: 10.15838/alt.2020.3.2.4.

7. Aksenov A.G., Teterin V.S., Ovchinnikov A.Yu. et al. Using a neural network to identify diseased potato plants. Agra­rian Science. 2022. N7-8. 167-171 (In Russian). DOI: 10.32634/0869-8155-2022-361-7-8-167-171.

8. Tsench Yu.S., Godlevskaya E.V. Mathematical modeling as a aspect for designing agricultural machines and units (development history of Southern Urals scientific school). Agricultural Machinery and Technologies. 2023. Vol. 17. N2. 4-12 (In Russian). DOI: 10.22314/2073-7599-2023-17-2-4-12.

9. Kalichkin V.K. On the need for a paradigm shift in agricultural research (message two). Siberian Herald of Agricultural Science. 2024. Vol. 54. N9(310). 102-115 (In Russian). DOI: 10.26898/0370-8799-2024-9-11.

10. Arshaghi A., Ashourian M., Ghabeli L. Potato diseases detection and classification using deep learning methods. Multimed Tools Appl. 2023. 82. 5725-5742 (In English). DOI: 10.1007/s11042-022-13390-1.

11. Fuentes A., Yoon S., Kim S.C., Park D.S. A deep robust-learning-based detector forreal-time tomato plant diseases and pests recognition. Sensors. 2017. Vol. 17. N9. 2022 (In English). DOI: 10.3390/s17092022.

12. Barbedo J.G.A. Factors influencing the use of deep learning for recognition of plant diseases. Biosystems Engineering. 2018. Vol. 172. 84-91 (In English). DOI: 10.1016/J.BIOSYSTEMSENG.2018.05.013.

13. Starovoitov S.I., Korotchenya V.M. A concept of digitalization of tillage machines. Machinery and Equipment for Rural Area. 2021. N8(290). 2-6 (In Russian). DOI: 10.33267/2072-9642-2021-8-2-6.

14. Lobachevskiy Ya.P., Lachuga Yu.F., Izmaylov A.Yu., Shogenov Yu.Kh. Scientific and technical achievements of agricultural engineering organizations in the context of digital transformation of agriculture. Machinery and Equipment for Rural Areas. 2023. N4(310). 2-5 (In Russian). DOI: 10.33267/2072-9642-2023-4-2-5.

15. Ivashova O.N., Gavrilovskaya N.V., Shchedrina E.V. Introduction of digital technologies to ensure the development of the agricultural industry. International Journal of Humanities and Natural Sciences. 2022. N3-2(66). 137-139 (In Russian). DOI: 10.24412/2500-1000-2022-3-2-137-139.

16. Derevyannykh E.A., Mitrofanova T.V., Sorokin S.S. et al. On the application of artificial intelligence in agriculture. Vestnik Chuvash State Agrarian University. 2023. N4(27). 182-187 (In Russian). DOI: 10.48612/vchd2ut-5bhh-4dkk.

17. Ovchinnikov A.Yu., Teterin V.S., Panferov N.S., Pekhnov S.A. Development of a system for assessing the three–dimensional position of an infected potato plant. Agrarian Scientific Journal. 2025. N3. 136-142 (In Russian). DOI: 10.28983/asj.y2025i3pp136-142.

18. Amit Yu., Felzenshvalb P., Girshik R. Object detection. Computer vision: a reference guide. Springer International Publishing. 2021. 875-883 (In English). DOI: 10.1007/978-3-030-63416-2.

19. Khanam R., Hussain M. YOLOv11: An overview of the key architectural enhancements. arXiv preprint arXiv. 2410.17725.2024 (In English). DOI: 10.48550/arXiv.2410.17725.

20. Jafar A., Bibi N., Naqvi R.A. et al. Revolutionizing agriculture with artificial intelligence: plant disease detection methods, applications, and their limitations. Frontiers in Plant Science. 2024. Vol. 15. 1356260 (In English). DOI: 10.3389/fpls.2024.1356260.