Harmful Insects Detection Using Convolutional Neural Networks (Faster R-CNN) *

International audience Insect detection is a crucial task in various fields, including agriculture, entomology, and biodiversity conservation. Among the problems we encountered was the difficulty of identifying insects due to the great similarity of appearance of certain species. Currently, the Convolutional Neural Networks (CNNs) have been widely adopted for insect detection due to their ability to accurately classify objects in images, using recent advances methods in deep learning and computer vision algorithms. In this paper, we focused only on seven types of insects most harmful to agricultural crops in Morocco, such as olive and wheat… We propose a CNNbased architecture specifically Faster RCNN to processing our model. The purpose of this research is to determine the type of insect and monitor it, which can allow us to identify and reduce the chemical pesticides used but also to take timely preventive measures and avoid economic losses.