Deep Learning-Based Detection of Foliar Diseases in Apple Plants Using an Assembled CNN Model

The apple industry faces significant economic losses due to diseases and pests, contributing to low productivity levels. Detecting apple diseases promptly is essential for controlling their spread and improving overall productivity. However, early identification and diagnosis of these diseases are challenging due to various factors. In this work, we evaluated the performance of three distinct models: a custom CNN, Xception, and Resnet152V2. These models were tested on an expert-annotated dataset of apple diseases, comprising approximately 1821 high-quality RGB images across different categories of following types: C1: scab affected apple, C2: cedar apple affected by rust, C3: diseases of complex nature, and C4 samples of healthy apples. The custom CNN based proposed model trained for 50 epochs could yield an approximate accuracy of around 95%.