MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641079707 A) filed by Sona College Of Technology on June 29, 2026, for Ai-Based Early Crop Disease Detection System For Tamilnadu Farmers With Tamil Language Voice Advisory.
Inventors include Ilanchezhian P; Shanmugaraja P; Vasanthi S; Thangaraj K; Aldo Stalin J; Rithish I; Harshavardhan G; and Harshanth G.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: ABSTRACT This invention introduces an AI-powered system to catch crop diseases early, built just for smallholder farmers in Tamil Nadu. Farmers snap a photo of a leaf with their regular Android phone, and within seconds, they hear a diagnosis in Tamil, straight from their device. Here's how it works. First, the system uses a YOLOv8n-seg model to spot and outline any diseased patches on the leaf and figures out how much of the leaf is affected. Next, a fine-tuned EfficientNetV2-S model checks the diseased area to pinpoint exactly which disease it is, aiming for over 93% accuracy. To make things transparent, the system overlays a Grad-CAM heatmap on the farmer's photo, so they can see where the AI focused its attention. Everything runs on a cloud server using FastAPI. Once a farmer sends in a picture, the server sends back a full report: the disease name, how confident the AI is, how severe the disease is, the Grad- CAM heatmap, and a treatment advisory all in Tamil and all in under three seconds. The mobile app is light, under 15 MB, and works with Android 7.0 or newer. It lets farmers take or upload leaf photos, shows the results with the heatmap, and plays a Tamil audio message generated by the AI4Bharat Indic TTS engine. This message covers both organic and chemical treatments, the right dosage, preventive tips, and even the estimated cost per acre. The team trained and tested the system using a custom, Tamil Nadu-specific dataset with over 20,000 annotated leaf images from five major crops and more than twenty disease types. They're also releasing everything: the dataset, model weights, and app code as open-source tools, so the whole Indian agricultural AI research community can benefit.
Disclaimer: Curated by HT Syndication.
