Authors: G. V. Rishitha, K. Sai Rohith, G. Gurumurthy, E. Raju, Mrs. Jahanara Begum
Abstract: Crop diseases significantly reduce agricultural productivity, and early detection is essential for preventing further damage. In practice, however, farmers largely depend on visual inspection of plants, a method that is time-consuming, error-prone and often unreliable without expert help, which is frequently unavailable in rural areas. This paper presents AgriScan AI, a web-based plant disease detection application that allows a farmer to upload or capture a photo of an affected plant leaf and receive an analysis within seconds. The proposed system uses YOLO26n, the nano model of the Ultralytics YOLO26 family, a lightweight Convolutional Neural Network (CNN)-based model, in its image-classification form (yolo26n-cls) fine-tuned on the PlantVillage dataset. The model analyses the leaf image and identifies whether the plant is healthy or diseased, and the application displays the possible disease name based on the model’s prediction together with a confidence value, a severity indication and guidance on cause, treatment and prevention. The application follows the workflow CAPTURE → UPLOAD → PREPROCESS → CLASSIFY →DISPLAY → ADVISE → RECORD. Beyond classification, the prototype provides farm-management features including field mapping, scan history, a weather summary, an assistant named Krishi Mitra, an outbreak map, a farm diary and expert-help access, all within a simple and user-friendly interface. The goal is to reduce the need for manual inspection and frequent expert visits and to make disease detection faster and more accessible in remote areas. The current prototype operates in demonstration mode with illustrative results; quantitative metrics reported in this paper are illustrative prototype values and should be replaced with measured results from the trained YOLO26n model before final submission.
DOI: https://doi.org/10.5281/zenodo.24135894
