Weed Identification in Crop Field Using CNN


Saniya Zahoor
Post Graduate Department of Computer Science, University of Kashmir, Hazratbal.
Shabir A.Sofi
Department of Information Technology, National Institute of Technology Srinagar.


Since the ages agriculture has remained as the backbone of economies especially developing countries like ours, where population is growing rapidly being second most populated country in the world, food demands are increasing so, farmers need to maximize their productivity. Weed is one of the enemies to farmer’s crop which competes with the crop for nutrients and sometimes hinders the growth of crop. Weed can cause loss of production ranging from 10 to 100%. There has been research on the use of many CNN models for weed identification. This paper presents a classification model to distinguish between weed and crop images and it classifies 12 species of weeds and crops. The proposed model achieves 96.45% of accuracy during training and of 90.08% during validation and testing.