Existing Systems

There are ways of looking at the current model which exists for detecting plant disease,

The traditional method, which is currently used in many parts of the world and is not a favorable method in our perspective. In this method the farmer has to take the soil samples and the leaf samples and submit it to the lab, where it gets thoroughly tested and after a while the result is submitted to the farmer. This method is accurate but takes a lot of time, during which the farmer can lose a decent amount of crops to a particular disease.

Proposed System

Here we proposed to use the Artificial Neural Network technique, to classify whether the plant is infected or not based on different parameters. Convolutional Neural Network (CNN) would be used to classify between the images thereby allowing the user to quickly identify the type of the disease if the plant is infected with one, the model will be trained on different datasets in the form of images to make the necessary predictions. In deep learning, Convolutional neural networks (CNNs) are one of the main categories which participate in images recognition and image classifications. CNN is widely used in objects detections, recognition of faces etc. Predominantly Python libraries such as TensorFlow, Keras, etc. would be used in the model, the transfer learning method will also help to size down the production process and increase the model accuracy.

Working on the Proposed System

In CNN, the image is taken in the form of a pixel and a random filter is generated which is then multiplied with the corresponding pixel and its neighboring pixels. A summation of these is taken and a second matrix is generated. In this, MaxPooling2D is performed and the matrix is divided into smaller divisions, the maximum pixel value of each of these smaller divisions is selected which then enhances the image. Data preprocessing is carried out in the first step of CNN and augmentations such as image resizing, reshaping is done. The image is converted in a form of an array which holds the pixels values. Then the image is processed through various layers which are encompassed in CNN which are the convolution2D layer, maxpooling2D layers, BatchNormalization, flatten layer, dense layer and the dropout Layer. After the model is trained successfully it can accurately classify the disease which the plant has been infected with.