Classification of Potato Leaf Disease using Deep Learning

Deep learning is one of the subdomains of Artificial Intelligence. With the advantages of automatic learning and feature extraction, in recent years, it has become a hot topic in academic and industrial circles. Meanwhile, this has emerged as one of the hotspots within agricultural plant protection research, including plant disease recognition and pest range assessment. Deep learning in plant disease recognition can avoid the shortcomings of artificial selection of features of the disease spot, making the feature extraction of plant diseases more objective, promoting research efficiency, and increasing technological transformation speed. The goal of this work is to analyze potato leaf diseases using CNN in order to determine the model’s optimal performance using a large dataset (10,000 photos) of potato leaves in the analysis. The most effective model will support farmers in making decisions that will keep potato production from declining. This investigation revealed that the model worked better when forty epochs were applied to the dataset of 10,000 photos. Additionally, three different epochs were applied and achieved 100% accuracy for 30 and 50 as achieved 99.97% and 99.98% respectively. By offering recommendations for how to categorise Potato Leaf Disease with the best result, the study made a substantial contribution to the agricultural industry and farmers. While working with big volumes of data, researchers should be concerned about the system because running code takes longer time. To focus this, a dataset including 2152 photos was gathered from Kaggle, with an additional 7848 images sourced from various platforms such as Google and raw data from Potato Field. To determine which epochs performed the best, a range of epochs were applied and results shows that 40 epochs led to 100% accuracy.