Apple Leaf Disease Classification Using Big Transfer with Modified Z-Score

Malus domestica, commonly referred to as apples, stands as the second most popular fruit globally, surpassing bananas in revenue generation both per hectare and overall. This nutrient-dense fruit boasts high levels of fiber, vitamin C, antioxidants like vitamin E, and polyphenols, offering a myriad of health benefits. These advantages include potential risk reduction for cancer, diabetes, and heart disease, as well as support for weight management, gastrointestinal health, and cognitive well-being. Notably, the presence of pectin, a specialized fiber, serves as a probiotic, promoting the growth of beneficial bacteria in the gastrointestinal tract. Despite its significant healing properties, apples are susceptible to various diseases such as black rot, cedar apple rust, and scabs, leading to substantial economic losses. This study proposes a deep-learning approach utilizing Big Transfer R50×1 with a custom head featuring a 9-layer design. The dataset used is a subset of the PlantVillage dataset, focusing on four classes: black rot, cedar apple rust, apple scab, and healthy, comprising a total of 4645 images. The model underwent training and evaluation, achieving an impressive accuracy score of 99.25% using 18 seconds per epoch. Precision, recall, and f1-score were also remarkable at 99.25%. The findings of this study confirm the efficacy of the proposed approach in precisely categorizing a range of apple diseases, alongside its expedited performance in classification endeavors. This positions the method as a significant and applicable asset within the realm of apple cultivation.