Using Deep Learning Models for Crop and Weed Classification at Early Stage

Abstract

Agriculture is essential for human existence, and it plays an important role in the world economy. There is increasing demand for food to feed the everincreasing world population. Agriculture is affected by climate changes along with weed control. Weeds are unwanted plants that compete with plants for nutrition, and sunlight and adversely affect crop quality and production. Manual weeding is a tedious and labor-intensive task because both crop and weed look the same by visual appearance. Artificial intelligence techniques like deep learning can address this problem of crop and weed classification. In this research work, a deep learningbased classification system has been proposed to classify the weed and crop based on RGB images. We investigated two popular deep learning-based transfer learning models, namely DenseNet169 and MobileNetV2, and assessed their performances for crop and weed recognition. These models perform excellently with an accuracy of 97.14 and 94.92%, respectively. The significant accuracy results make the model an important tool for farmers to identify weeds.