Agricultural crop disease classification has been performed with the images from PlantVillage dataset using pre trained deep learning architecture namely VGG16 net.
The model was implemented with callbacks - Early Stopping, Reduced Learning Rate On Plateau and Model checkpoints. The classification accuracy using 70295 images was approximately 95% for VGG16 net .
The performance of the models has been evaluated by modifying the number of images, setting various batch sizes and varying the weight and bias learning rate. The number of images significantly affected the performance of the model.
With the crop diseases being correctly identified minimum 95% of the time, we can aspire to help our farmers detect crop infections at an early stage. This would aid them to plan farming techniques that are more effective than the previous ones they were following.
