Hybrid Vision Mamba Transformer Network for Accurate Tomato Leaf Disease Classification

Authors

  • Rabinarayan Panda Author
  • Sachikanta Dash Author
  • Dharamvir Author
  • Mahaboob Basha M.A Author
  • Bishnu Paramguru Mahapatra Author

Keywords:

Tomato leaf disease classification, Vision Mamba, Vision Transformer, Deep learning, Precision agriculture, PlantVillage dataset, Hybrid neural network.

Abstract

Timely identification of tomato leaf diseases can help farmers preserve yield and achieve better field management. Deep image models are able to address this problem. Convolutional neural networks perform great on local relations, while they fail to model relationships between distant regions of an image. On the other hand, Vision Transformers allow capturing the large context; however, they are quite costly in terms of computational expenses.In this study, we propose a Hybrid Vision Mamba Transformer Network, referred to as HVMTNet, to automate the tomato leaf disease detection process. The architecture utilizes various components. At first, a conv stem learns local features from an image. Next, a Vision Mamba encoder is applied to model long-range dependencies efficiently. Afterward, a Transformer encoder is used to learn broader scene context. Finally, adaptive feature fusion is performed to aggregate the learned cues and generate disease-specific features.HVMTNetis evaluated on the PlantVillage tomato leaf disease dataset. There are ten disease classes in it. The data are partitioned into subsets using 70:15:15 ratio. The proposed approach outperforms ResNet50, EfficientNet-B0, ConvNeXt-Tiny, Vision Transformer, and Vision Mamba models. The obtained results are as follows: 99.61% accuracy, 99.54% precision, 99.48% recall, 99.51% F1-score, and AUC 0.999.

Downloads

Download data is not yet available.

Downloads

Published

2026-09-22

How to Cite

Hybrid Vision Mamba Transformer Network for Accurate Tomato Leaf Disease Classification . (2026). Minimax Theory and Its Applications, 11(2s), 849-866. https://journalmta.com/index.php/jmta/article/view/458