A DATASET OF FIELD PLANT IMAGES FOR PLANT DISEASE DETECTION AND CLASSIFICATION WITH DEEP LEARNING

A DATASET OF FIELD PLANT IMAGES FOR PLANT DISEASE DETECTION AND CLASSIFICATION WITH DEEP LEARNING

Dr.N.Gopal Krishana, Vegi Sandhya Rani, Syed Shaheen, Sarimalla Anusha, Ponugubati Rajani.

Computer Science and Engineering

Tirumala Engineering College, Narasaraopeta

Abstract - The United Nations Food and Agriculture Organization proposes a 70% increase in food production by 2050 to meet global population demands, recognizing that one-third of all food is lost due to plant diseases. To address this challenge, researchers have developed deep learning models to aid farmers in detecting crop illnesses. While these models are typically trained on databases like Plant Village or Plant Doc, which contain laboratory photos with consistent backgrounds, they perform poorly when applied to field images with complex backgrounds and multiple leaves per image. To remedy this, a new dataset called Field Plant has been introduced, comprising 5,170 plant disease images collected directly from plantations and manually annotated by pathologists. This dataset focuses on diseases affecting crops like corn, cassava, and tomato in tropical regions. Evaluation of state-of-the-art models on Field Plant shows superior performance compared to other datasets. Additionally, recommendations on crop cultivation are made based on factors such as soil quality, weather conditions, humidity, and rainfall, aimed at increasing agricultural output. This system benefits farmers, nations, and helps to stabilize food costs.

Keywords - Boosting food production,Deep learning models,Plant diseases,Field Plant dataset,Soil, weather, humidity, rainfall,Agricultural output.