Early Detection and Classification of Multi Crop Leaf Diseases using Custom Based CNN Model

Plant monitoring and disease diagnostics are essential to sustainable agriculture. However, plant diseases are difficult to manually monitor since they require accurate and timely diagnosis. In this paper, we propose Early Detection and Classification of Multi Crop Leaf Diseases using Custom Based Convolution Neural Network (CNN) Model. The images from standard Kaggle dataset are subjected to preprocessing which includes Resizing, Adaptive Histogram Equalization, Smoothening and Banalization. The preprocessed data is subjected to custom based CNN to obtain the high level hidden features. The proposed model is trained on data samples for different diseases with dataset types of Mango, Gauva and Chilli Leaves. This achieved average accuracy of 97.65% and precision of $\mathbf{9 6. 3 5 \%}$ for 3858 image samples. The experiment findings demonstrate that the proposed method offer a more effective way to control plant diseases.