RECOGNITION OF CROP DISEASE AND INSECT PESTS BASED ON DEEP LEARNING IN HARSH ENVIRONMENT

Abstract

One of the most significant variables that significantly endangers agricultural output is the 
presence of agricultural diseases and insect pests. Eliminating pest-related economic losses is 
possible via early detection and identification. This research presents an approach to 
automatically identifying crop diseases using convolution neural networks. Each of the ten crops 
included in the dataset has 27 photos of diseases; the dataset is sourced from the 2018 AI 
Challenger Competition's public data set. The Inception-ResNet-v2 model is trained in this work. 
Two components of the model's residual network—the cross-layer direct edge and the multi- 
layer convolution. Once the combined convolution procedure is finished, the ReLu function is 
called upon to activate it. The total recognition accuracy in this model is 86.1%, according to the 
trial findings, proving its usefulness. We created a Wechat applet that can identify agricultural 
illnesses and insect pests after training this model. After that, we administered the exam itself. 
The results demonstrate the system's ability to correctly detect crop illnesses and provide 
appropriate recommendations.