Year: 2025
Type: article
Abstract: Crop diseases significantly reduce agricultural yields and present a challenge to food security and economic development. Traditional disease detection methods are manual, subjective, and often unavailable in remote regions. This project develops a low-cost AI-based crop disease detection system using MobileNetV2 CNN deployed on a Raspberry Pi platform. The system uses TensorFlow Lite with INT8 quantization for efficient on-device inference. It targets the detection of 14 disease classes across three important crops: potato, tomato, and corn utilizing the PlantVillage dataset augmented with field images. Operating entirely offline with a Python-based GUI, the system offers a practical tool for farmers at a hardware cost of ₹17,550. The system aims for 94%+ accuracy and sub-two-second latency for on-field deployment. The system identifies diseases in potato, tomato, and corn crops across 14 disease classes. Operating completely offline with a user-friendly touchscreen interface built with Python and Tkinter, it provides instant diagnosis and treatment recommendations. The hardware cost is maintained at ₹17,550, making it accessible for small and medium farmers.
Authors: Pawankumar Bari, Om Kulkarni, Prathamesh Kudale, Shrushti Hadole, Sujata Mali
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
Open Access status: gold
OpenAlex ID: W4416061626