Frequency-Aware and Residual-Attention CGAN with Two-Stage Training for Imbalanced Potato Leaf Disease Detection in Crop Management

Key Points

• The system detects potato leaf diseases effectively, enhancing crop management outcomes.
• Key evidence shows that leveraging residual-attention and a two-stage approach improves accuracy by 35%.
• Analysis utilizing a frequency-aware conditional generative adversarial network enhances detection precision.
• The method highlights the need for improved training techniques to address imbalanced data challenges.