A novel approach for insect-pest identification using multipath convolutional neural network

For sustainable agriculture development and ensuring food security, there is an urgent need to improve the detection, monitoring, prediction, and identification of crop diseases and insects. Every year about 37% of rice crop gets damaged due to pests and insects. Without using any preventive measures, we could have lost about 70% of crops due to pests and insects. This study aims to identify the insects at the early stage using a multipath Convolution Neural Network. The main objective of this proposed work is to devise a model to efficiently identify the pest by images captured by mobile cameras. At the initial stage, it basically focuses on the classification of 6 classes of insects. The accuracy of classification for different classes of crop pests and insects achieved was 99%. The proposed method’s accuracy is better than CNN, Faster CNN, DenseNet-121, and Deep Residual Learning.