RESNEXT-BASED IMAGE CLASSIFICATION MODEL FOR PLANT DISEASE IDENTIFICATION
Advancement in Automation has brought a revolution in the field of agriculture for various applications. Digital image processing combined with Machine Learning techniques can provide support in the area of agriculture by assisting in the classification and detection of plant diseases. In this paper, the ResNext-50 model is presented for the classification of plant diseases and has proved to be more efficient not only in terms of Image Classification but also in terms of Image Segmentation. The model is developed by making a few modifications to the ResNet-50 model and adding some layers. A dataset from Kaggle is used having 87867 images for training and validation. ResNet-50 and ResNext-50 are trained on 25 epochs with 80% images and then validation is done with the remaining 20% images. The ResNext-50 model is found to have better training and validation accuracy than the ResNet-50 model. ResNext-50 model is found to have a training accuracy of 99.65% and a validation accuracy of 94.05% after the last epoch, on the other hand, the ResNet-50 model is found to have a training accuracy of 99.32% and a validation accuracy of 93.45% after the last epoch. The accuracy and loss results have proved the efficiency of the ResNext-50 model over the ResNet-50 model.
CNN, Deep Learning, Machine Learning, Plant Disease Classification, ResNext Model.