Weed Seedling Detection Using Mask Regional Convolutional Neural Network

Abstract

Agricultural production is affected by many factors; in which weed is one of them to reduce the production by soaking nutrition of the original plants. There are several traditional methods such as pesticides, herbicides to remove the weeds but they affect the soil fertility, which ultimately affects the production. The proposed work focuses on the state-of-the-art technologies such as machine learning to smartly detect the weeds and help to provide some eco-friendly techniques to remove them as well. This paper proposes an enhanced "Mask R CNN" model that is aimed at extracting images from `cranesbill seedling dataset' to detect cranesbill seedlings during their early stage. Besides, weed is also beneficial and used as an herbal medicine for the treatment of rheumatism and headaches. The proposed research work gives an idea on how Mask Regional Convolutional Neural Network helps to segregate weed (cranesbill) from the original crop so that crop gets complete nutrition and increases the production. The proposed model has achieved a hundred percentaccuracy in the training phase and more than ninety-eight percent accuracy in the validation phase.