Human visual observation and inspection are the most common methods of diagnosing cotton plant diseases. In addition to taking a long time and requiring considerable effort on the part of the researcher, it requires specialized knowledge and an extensive amount of time to complete. As a result of recent advancements in computer vision and deep learning, a system for the detection of plant diseases may be developed in which disease can be detected across a wide geographical area with a minimum of intervention on the part of humans. A variety of neural network algorithms have been evaluated across a variety of experiments, such as Mini batch gradient descent and Faster RCNN. A total of 54305 images were used in all of the experiments, representing 38 different classes of plant diseases. The performances of each architecture were evaluated using four different metrics: accuracy, precision, recall, and F1 score. Compared with other convolutional neural network (CNN) architectures, our proposed model performed better than all other Convolutional Neural Network (CNN) architectures by an average of 95.12% and 96.98% respectively for 6 and 40 training epochs.

Cotton Plant Disease, Detection, Classification, Mini Batch Gradient Descent, Faster RCNN.