Deep Learning Model For Weeds Detection & Classification

Abstract

Hunger remains a persistent and serious global issue affecting millions of people worldwide. Despite advancements in agriculture and food distribution, chronic hunger and malnutrition continue to plague communities and nations. The rise in global population further amplifies the challenge of feeding people adequately. Innovativsolutions are being sought, and deep learning techniques offer promising avenues for addressing agricultural problems, particularly weed eradication, which poses a significant threat to crop production. Smart farming, empowered by advanced technologies like artificial intelligence (AI), presents a more efficient, precise, and sustainable approach compared to traditional agriculture. In this work, we present an initial effort towards a smart farming solution for weed eradication. Our approach applies transfer learning on DenseNet121 a Deep Convolutional Neural Network (CNN) pretrained on imagenet, trained on a dataset comprising images of eight weed species and various flora. The goal is to detect and classify weed images accurately, serving as a crucial first step towards developing robotic systems that can be deployed in agricultural fields. By harnessing the power of deep learning, we aim to contribute to the development of effective and automated weed eradication strategies. Despite not contributing much, this research holds significant potential to alleviate the challenges posed by weeds in agriculture and advance the adoption of smart farming practices. Keywords: Deep Learning, Convolutional Neural Network, Weeds Images, Images Classification.