MUMBAI, 23 June 2025: In a significant advancement for smart agriculture in India, researchers from IIT Patna, IIT Bombay, and the Rajiv Gandhi Institute of Petroleum Technology, Amethi, have developed EdgePlantNet—a real-time, AI-driven system capable of detecting plant diseases using affordable edge computing devices such as the Raspberry Pi.
Designed to empower farmers with rapid insights, EdgePlantNet tackles a core challenge in agriculture: identifying plant diseases early and accurately without relying on heavy chemical usage or cloud-based processing that can cause delays.
“India's dependence on healthy crops makes early disease detection critical,” the researchers noted. Traditional pesticide approaches are often broad-spectrum and environmentally harmful. EdgePlantNet, however, uses Convolutional Neural Networks (CNNs) and a novel MLP-based spatial attention mechanism (MLP-ATCNN) to analyze leaf images directly on the field.
What sets the system apart is its dual-image input mechanism. It simultaneously processes an original photo of the leaf and a segmented version highlighting diseased spots. This allows the system to learn more accurate patterns using k-means clustering to isolate non-green areas for disease detection.
Despite its advanced capabilities, EdgePlantNet is incredibly lightweight, boasting fewer than 200,000 parameters, with the attention mechanism comprising less than 5,000. It runs efficiently on a Raspberry Pi, processing images at 3.65 frames per second and consuming just 4MB of memory—far smaller than other AI models like ResNet or DenseNet.
In tests using the PlantVillage and BPLD datasets, EdgePlantNet achieved standout accuracy:
- 99.2% for potato leaves,
- 97.1% for tomato leaves, and
- 95.7% for black gram in more natural conditions.
It was especially effective at detecting “few-shot diseases”—those with very few training examples—demonstrating real-world applicability in rural environments where unknown pathogens may emerge.
This advancement reflects the next frontier of precision agriculture—bringing AI and IoT tools out of the lab and into the hands of farmers. With EdgePlantNet, even small farms can now detect issues early, reduce pesticide use, and boost crop yields without costly infrastructure.
The research team believes the tool can be further improved to detect multiple diseases on a single leaf and handle more extreme environmental conditions. For now, EdgePlantNet marks a major step forward in democratizing AI-driven farming tools for a greener, more sustainable future.
