An Agentic Hybrid EdgeMambaViT-CNN Framework for Real-Time Mango Leaf Disease Detection

DOI

10.2991/978-94-6239-789-7_27How to use a DOI?

Keywords

Mango leaf disease detection; Deep learning; Vision Transformer; Mamba architecture; Agentic AI; Edge computing; Precision agriculture

Abstract

Mango leaf diseases have a significant effect on crop yield and quality, and it presents a significant challenge to sustainable agriculture. These diseases need to be detected early and accurately to reduce the losses and aid in the proper management of the disease. This paper presents a new Agentic Hybrid EdgeMambaViT-CNN architecture to ensure mango leaf disease detection in real time and robustly. The proposed model combines CNN to extract local features, Vision Transformers (ViT) to learn the global contextual features, and a Mamba-based sequence model to learn long-range dependencies with lower computational complexity. Moreover, an attention-based feature fusion mechanism with the Convolutional Block Attention Module (CBAM) improves the ability of the model to give attention to disease-related parts. In order to further enhance adaptability and usability, an intelligent agent-based decision layer is integrated, which includes detection, diagnosis and continuous learning sub-elements. The system measures prediction confidence and dynamically adjusts the model with uncertain samples, thus improving performance in the long run. The model has been optimized to be implemented on the edge with TensorFlow Lite in order to allow real-time inference on mobile devices. The application of the proposed method to multi-class mango leaf disease samples shows that it is effective, with an overall accuracy of about 99, high precision, recall and F1-scores in all classes. The robustness and generalization ability of the model is confirmed by the confusion matrix, ROC, and the Precision-Recall curves. The suggested structure is an efficient, practical and scalable framework to be used in detecting intelligent plant diseases in real-world agricultural settings.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Vivek Kumar
AU  - Satender Kumar
PY  - 2026
DA  - 2026/10/08
TI  - An Agentic Hybrid EdgeMambaViT-CNN Framework for Real-Time Mango Leaf Disease Detection
BT  - Proceedings of theInternational Conference on Innovation and Practices in Artificial Intelligence (ICIPAI 2026)
PB  - Atlantis Press
SP  - 394
EP  - 402
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-789-7_27
DO  - 10.2991/978-94-6239-789-7_27
ID  - Kumar2026
ER  -