A Deep Learning Approach to Classify the Potato Leaf Disease

Purpose: This paper aims to classify potato disease using convolutional neural network in different epochs to observe the best performance of the model. The best model will help the farmers to make different decisions to prevent the loss of potato production.

Methodology: The paper implements a deep learning approach, the convolutional neural network, to explore potato disease classification. To accomplish the research objective, we collected 10000 images of potato leaves from different sources like google and raw data from potato fields. We collected a dataset of 2152 images from Kaggle and the other 7848 images from the above sources. The dataset belongs to a few classes. The classes are Potato Early Blight, Potato Late Blight, and Potato healthy leaf. The paper includes four main steps: data acquisition, data pre-processing, data augmentation, and image classification to find the output.

Findings: This study found that the model performed better when we applied 40 epochs for the 10000 images dataset & we achieved 100% accuracy as we applied a total of 3 different epochs and achieved an accuracy of 99.97% and 99.98% for 30 and 50 epochs, respectively.

Research Limitations: The study significantly contributed to the agriculture sector and farmers by providing suggestions to classify the Potato leaf Disease with the best output.

Besides, researchers need more raw data to build the model for better output, and they also should be concerned regarding the system when working with large volumes of data as it takes longer to run the code.
Originality/Value: This research paper contained high volume of the dataset, which is 10000 images of potato leaves. We collected a dataset of 2152 images from Kaggle and the rest, 7848 images from different sources like google, and raw data from potato filed. We showed different epochs to check the best performance and achieved 100% accuracy when 40 epochs were applied.