Attention mechanisms and transfer learning for robust peach leaf damage classification under domain shift
Researchers have developed a new deep learning model for classifying peach leaf damage, achieving high accuracy on a benchmark dataset. The model, an enhanced EfficientNetB5 incorporating a Convolutional Block Attention Module (CBAM), reached 93.3% accuracy. Transfer learning strategies were then applied to adapt the model for real-world conditions, with an attention-enhanced EfficientNetB3 achieving a 93% macro F1-score on a local dataset, demonstrating improved robustness and generalization. AI
IMPACT Enhances AI's utility in agriculture by improving automated crop damage assessment and decision-making.