A Comprehensive Analysis of Multi-Fruit Disease Recognition Through Deep Learning and Traditional Machine Learning Modeling

Rapid, accurate fruit-disease detection boosts yields for high-value crops such as mango, guava, citrus and pomegranate, yet manual inspection is slow and error-prone. We therefore built an image-based system around a bespoke CNN with extra convolutional layers. Trained on 30,260 pictures spanning 17 diseases\u2014from citrus black spot to pomegranate bacterial blight\u2014the network reached 99 % validation accuracy and surpassed ResNet-50, VGG-16, MobileNet-V3 and conventional algorithms (k-NN, SVM, Random Forest, Decision Tree). The model now powers a lightweight web app that gives farmers real-time diagnoses.