An experimental set up for utilizing convolutional neural network in automated weed detection

Weeds present in the crops are one of the factors that lead to decrease in crop production. The weeds take up nutrients, water which leads to weight reduction in plant and decreases the grains per ear and grain yield.so a method needs to be developed which would detect these weeds in the field and then herbicide is sprayed on them to completely destroy through the use of new drone technology and deep learning in field of convolutional neural networks. Using the dataset from the fields of Indian Agriculture Research Institute (IARI), THE AUTHORS have applied transfer learning technique to identify three weeds Phalaris minor, Dactyloctenium aegyptium, Digera arvensis, Echinochloa colona which achieved achieve accuracy of 95%, 65%, 61%, 54% respectively for the datasets which were not involved in training of model.