Plant Disease Detection and Treatment System for Home Gardeners Using MobileNetV2 Transfer Learning

This paper presented PDDTS, a treatment-first plant disease detection system aimed squarely at home gardeners and plant enthusiasts, a user group that existing plant-disease research largely overlooks in favour of commercial-scale agriculture and bare classification labels. A MobileNetV2-based transfer-learning classifier, trained on a small, imbalanced seven-class subset of PlantVillage with balanced class weighting, reached 98.28% validation accuracy and a per-class F1-score of at least 0.96 while remaining light enough (2.6 million parameters) for interactive web inference without a dedicated GPU. More importantly, every prediction is carried through to a structured PostgreSQL treatment-reference table — covering severity, chemical treatment, organic treatment and prevention — and paired with live OpenWeatherMap data, so that a diagnosis always arrives with an actionable next step and relevant environmental context, delivered through Detect, Result, History and Dashboard views in a Flask web application. Confusion-matrix analysis and an end-to-end walkthrough of the deployed screens support the conclusion that a compact CNN combined with a well-structured treatment knowledge base can give home gardeners fast, trustworthy, near-actionable plant-care support. Limitations: Two limitations are worth stating plainly. First, like most PlantVillage-derived work, the training images are largely lab-condition photographs against clean backgrounds, and Section 2 notes that field-condition accuracy typically drops relative to lab-condition accuracy [2]; performance on genuinely cluttered home-garden photographs (uneven lighting, soil or pot in frame, partially occluded leaves) has not been separately measured. Second, the species and disease coverage is deliberately narrow — three plants and seven classes — so a gardener growing a species outside this set receives no diagnosis at all rather than a degraded one. Both limitations are addressed directly in the future work below.

[1] S. P. Mohanty, D. P. Hughes, and M. Salathé, “Using Deep Learning for Image-Based Plant Disease Detection,” Frontiers in Plant Science, vol. 7, Art. 1419, 2016. [2] K. P. Ferentinos, “Deep Learning Models for Plant Disease Detection and Diagnosis,” Computers and Electronics in Agriculture, vol. 145, pp. 311–318, 2018. [3] E. C. Too, L. Yujian, S. Njuki, and L. Yingchun, “A Comparative Study of Fine-Tuning Deep Learning Models for Plant Disease Identification,” Computers and Electronics in Agriculture, vol. 161, pp. 272–279, 2019. [4] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” in Proc. IEEE CVPR, pp. 4510–4520, 2018. [5] D. P. Hughes and M. Salathé, “An Open Access Repository of Images on Plant Health to Enable the Development of Mobile Disease Diagnostics,” arXiv preprint arXiv:1511.08060, 2015. [6] A. Kamilaris and F. X. Prenafeta-Boldú, “Deep Learning in Agriculture: A Survey,” Computers and Electronics in Agriculture, vol. 147, pp. 70–90, 2018. [7] A. K. Rangarajan, R. Purushothaman, and A. Ramesh, “Tomato Crop Disease Classification Using Pre-Trained Deep Learning Algorithm,” Procedia Computer Science, vol. 133, pp. 1040–1047, 2018. [8] F. Chollet et al., “Keras,” https://keras.io, 2015. [9] PostgreSQL Global Development Group, “PostgreSQL Documentation,” https://www.postgresql.org/docs/. [10] OpenWeather, “OpenWeatherMap Current Weather Data API Documentation,” https://openweathermap.org/current.