Abstract
Pearl millet is a staple food crop in areas with drought, low soil fertility, and
higher temperatures. Fifty percent is the share of pearl millet in global millet
production. Numerous types of diseases like Blast, Rust, Bacterial blight, etc., are
targeting the leaves of the pearl millet crop at an alarming rate, resulting in reduced
yield and poor production quality. Every disease could have distinctive remedies, so,
wrong detection can result in incorrect corrective actions. Automatic detection of crop
fitness with the use of images enables taking well-timed action to improve yield and in
the meantime bring down input charges. Deep learning techniques, especially
convolutional neural networks (CNN), have made huge progress in image processing
these days. CNNs have been used in identifying and classifying different diseases
across many crops. We lack any such work in the pearl millet crop. So, to detect pearl
millet crop diseases with great confidence, we used CNN to construct a model in this
paper. Neural network models use automatic function retrieval to help in classify the
input image into the respective disease classes. Our model outcomes are very
encouraging, as we realized an accuracy of 98.08% by classifying images of pearl
millet leaves into two different categories namely: Rust and Blast.