Research article

TOMATO LEAF DISEASE DETECTION USING CUTTING-EDGE DEEP LEARNING ARCHITECTURES

Alampally Sreedevi, Dr. Chiranjeevi Manike

Online First: May 25, 2023


Agriculture is vital to India’s economic progress. The major goal of agriculture the cultivation of a wide variety of valuable and necessary crops. Food safety could be adversely affected by plant diseases and generate substantial losses in production of agricultural products. For agriculture’s long-term viability, disease diagnosis on the leaf is critical. Due to time constraints and the complexity of disease, it’s difficult to make sense of plant diseases by hand. In the field of agricultural inputs, automatic classification of crop disease is widely required. Instance disease detection in plants is essential so it reduces the work that takes a long time of monitoring big farms and detects diseases at an early stage, limiting further plant degradation. In precision agriculture, deep learning has significantly contributed to classification and detection tasks. However, widespread adaption of these approaches and methodologies via low-cost limited devices for use in agricultural crops on a regular schedule is critical. Using the plant village dataset, which contains 87867 images divided into 38 classes to train and evaluate three deep neural network approaches for leaf disease classification. The detection accuracy of the 5-layer CNN, ResNet152 and EfficientNet-B3 architectures is about 88.32%, 93% and 97% respectively.

Keywords

Neural networks, CNN, deep learning, ResNet152, EfficientNet-B3