A Real-Time Application-Based Convolutional Neural Network Approach For Mango Leaf Disease Classification

The mango tree, thriving in tropical and subtropical climates, bears not only delicious fruits but also serves as a significant source of income. This study’s primary aim is to bolster mango production in Bangladesh, addressing the mango leaf diseases impacting mango yields. This study has proposed a Custom Lightweight Convolutional Neural Network (CLCNN) tailored to accurately classify seven distinct mango leaf diseases alongside their healthy counterparts. The CLCNN model, designed to offer an effective yet lightweight solution, emerges as a promising approach for disease classification. To affirm its efficacy, the performance of the CLCNN is compared against established pre-trained models such as VGG16, InceptionV3, F-Net, AlexNet, and ViT. The comparative analysis underscores the superiority of the proposed CLCNN model, attaining a notable testing accuracy of ${9 8 \%}$ surpassing the performance of pre-trained models. Moreover, the model is converted to TensorFlow light model which has been leveraged to develop an Android-based application for efficient classification of mango leaf diseases.