Search papers, labs, and topics across Lattice.
This study introduces an adaptive soft Mixture-of-Experts (MoE) framework that leverages cross-architectural routing among EfficientNet-B0, DenseNet-121, and Swin-Tiny to enhance plant leaf disease classification. By employing a soft gating mechanism for input-dependent expert weight assignment and a two-stage refinement training strategy, the model effectively captures both local and global features while addressing severe class imbalances. The proposed approach achieves notable performance improvements, with F1-scores exceeding those of the strongest individual expert by over 5%, demonstrating its efficacy across multiple datasets for real-world applications in crop health monitoring.
Achieving over 92% F1-score in plant leaf disease classification, this framework outperforms traditional single-architecture models by effectively integrating diverse feature representations.
Plant leaf disease classification is crucial for crop protection and precision agriculture but remains challenging under complex backgrounds, illumination variations, and severe class imbalance. Moreover, single-architecture models often fail to effectively capture both local and global representations. To address these challenges, this study proposes an adaptive soft Mixture-of-Experts (MoE) framework with cross-architectural routing that integrates EfficientNet-B0, DenseNet-121, and Swin-Tiny to exploit complementary multi-scale, local, and global features. A soft gating mechanism dynamically assigns input-dependent expert weights, while a two-stage refinement training strategy improves optimization stability and generalization. Experiments on a highly imbalanced potato leaf disease dataset achieve 91.68% recall and 92.62% F1-score, surpassing the strongest individual expert by 5.91% and 5.03%, respectively. Additional evaluations on durian and sesame leaf disease datasets yield F1-scores of 94.03% and 97.04%, demonstrating robust cross-dataset generalization and the potential of the proposed framework for reliable real-world crop health monitoring
