Abstract
Agriculture is vital to human survival and has a significant impact on the
economy of any nation. Crop protection costs millions of dollars per year. Insects and
other pests pose a serious threat to the health of a harvest. Excessive use of chemical
fertilizers and pesticides negatively affects the crop and soil quality. Therefore, one
way to safeguard the harvest and mitigate potential losses is through early
identification of the pests. Examining the crop at the right moment is the best technique
to determine its overall health. While manual inspection is the standard way of
conducting field inspection, it becomes challenging for large fields. In addition, manual
inspection would be exceedingly expensive and tedious. To address this, an automated
system is needed to detect pests, identify them, and recommend appropriate fertilizers
using an IoT system. Therefore, automated pest detection has become a major focus for
researchers globally, as it offers a more efficient and cost-effective alternative to
manual inspection. In this work, a smart agriculture system has been proposed that
monitors crops, identifies pests, and allows remote control. The dataset comprises over
4000 images of corn leaves, categorized into rust, blight, grey spots, and healthy
leaves. By employing Convolutional Neural Networks (CNN), the system has achieved
a remarkable 99% accuracy in pest detection.