| Popis: |
From the farmers’ perspective, it is very important to detect the beneficial and non-beneficial paddy pests for reducing the use of pesticides with increased productivity. In this chapter, a region-based deep convolutional neural network (Faster R-CNN) where a deep CNN is connected with a region proposal network (RPN) is used to perform the detection and identification paddy pests from the images. The main focus of this chapter is to find out not only harmful pests but also beneficial pests of paddy field so that farmers can make decisions when to use pesticides and when not to. A dataset with a large number of beneficial and harmful pests’ images from the paddy field was created for experimentation. Three models of Faster R-CNN based on ResNet-101, VGG-16, and MobileNet have been applied, and it is found that ResNet-101 gives the highest accuracy of 96.93% for beneficial and 95.07% for non-beneficial pest detection and identification compared to other networks. |
