Abstract:
Many tasks are difficult for us but easier for machine to do as to detect any
disease in plants it is difficult for human beings to find the diseases in plants without
heaving years of experience in farming which can cause immense effect to the
plants. In agriculture it is very important to recognize and find the disease of the
plants in the early stages. As, disease in the plants can affect the yield of the crops. It
is unhealthy for the plants and in return it can affect a lot to the farmer and in last
danger to the food-security. Using computer vision techniques, we can classify the
plants with the help of state-of the art ML algorithms and deep learning models to
differentiate between healthy and the effected plants by classifying their leaves. It is
one of the techniques that different researchers worked on different plants using
different techniques and different pre-trained deep learning networks (DenseNet121,
EfficientNetB0, InceptionV3, VGG19, and Xception) and classic machine learning
algorithms to detect the diseases on the plants using the plant leaves.
This research worked on two different approaches. First, the experimental
component trains models using different plants images. Then, the training of the
models that have already been trained. According to the results that we get in our
experiments in our work, the Xception networks performed admirably from deep
learning models. It gives us an accuracy of 89.93%. On other hand, from machine
learning algorithms Random Forest performed much better than the SVM and
Decision Tree. In addition to our results, it indicates that the performance of the pre trained network system gives the best findings for Plant disease detection.
