Authors:
Snigdha Zaman, Abdullah Hafez Nur, Rejwan Bin Sulaiman

Addresses:
Department of Computer Science and Engineering, International Islamic University Chittagong, Chittagong, Chattogram, Bangladesh. Department of Electronics Engineering, Alqedairi Group, Riyadh, Saudi Arabia. Department of Computer and Information Sciences, Northumbria University, Newcastle, Tyne and Wear, United Kingdom.

Abstract:

Plantation disease surveillance is still a major problem for the agriculture industry, especially for expensive products like mangos. Crop yield, quality, and farmer profitability can all be significantly increased by early and precise detection of mango leaf diseases. However, due to high expert expenses, a shortage of specialists, and inconsistent visual symptoms, manual diagnosis is frequently incorrect. This study presents an improved deep learning-based framework for automated categorisation of mango leaf diseases to overcome these drawbacks. Several cutting-edge deep learning architectures, such as VGG16, DenseNet201, MobileNet, MobileNetV2, and two recently released ensemble models, LeafNet1 (MobileNetV2 + DenseNet201) and LeafNet2 (VGG16 + DenseNet201), are used in this study. These models were chosen for their exceptional feature extraction and their ability to handle intricate spatial patterns in photos of damaged leaves. The models that performed well were VGG16 (98%), DenseNet201 (89%), MobileNet (98%), MobileNetV2 (88%), LeafNet1 (95%), LeafNet2 (99%), and VGG22 (79%). Notably, LeafNet2 outperformed all individual architectures, achieving a maximum accuracy of 99%, highlighting the value of ensemble approaches in enhancing classification resilience. For the early diagnosis of mango leaf disease, the suggested approach provides a quick, precise, and scalable solution. Our solution outperforms current deep learning-based detection techniques and greatly decreases reliance on human expertise by automating disease diagnosis. This approach enhances farmers' and stakeholders' decision-making capacity and provides a reliable tool for managing contemporary agricultural diseases.

Keywords: Mango Fruit; Deep Learning; Leaf Disease and Plantation Disease; Agriculture Industry; Deep Learning Architectures; Visual Symptoms; Ensemble Approaches; Detection Techniques; Disease Diagnosis.

Received on: 12/06/2025, Revised on: 17/08/2025, Accepted on: 28/10/2025, Published on: 18/08/2026

DOI: 10.69888/FTSHSL.2026.000729

FMDB Transactions on Sustainable Health Science Letters, 2026 Vol. 4 No. 3, Pages: 154-166

• Views : 101
• Downloads : 10

Download PDF