Identification of Plant-Leaf Diseases Using CNN and Transfer-Learning Approach

The timely identification and early prevention of crop diseases are essential for improving production. In this paper, deep convolutional-neural-network (CNN) models implemented to identify diagnose in plants from their leaves, since CNNs have achieved impressive results the field machine vision. Standard CNN require a large number parameters higher computation cost. we replaced standard convolution with depth=separable convolution, which reduces parameter were trained an open dataset consisting 14 different plant species, 38 categorical disease classes healthy leaves. To evaluate performance models, such as batch size, dropout, numbers epochs incorporated. disease-classification accuracy rates 98.42%, 99.11%, 97.02%, 99.56% using InceptionV3, InceptionResNetV2, MobileNetV2, EfficientNetB0, respectively, greater than that traditional handcrafted-feature-based approaches. comparison other deep-learning model better terms it required less training time. Moreover, MobileNetV2 architecture is compatible mobile devices optimized parameter. showed promising can greatly impact efficient diseases, may potential detection real-time agricultural systems.