SwinLeafExplain: Interpretable Cotton Disease Classification with SWIN Transformer

Healthy agriculture is vital for a nation’s prosperity. Cotton is a significant cash crop that is vulnerable to several leaf diseases that can lower yield, such as powdery mildew, army worm, bacterial blight, target spot, and aphids. To prevent further damage to crops, it is essential to discover these leaf diseases early. The early detection of cotton leaf disease is greatly aided by advancements in artificial intelligence. SwinLeafExplain, a novel approach for cotton leaf disease classification, leverages a SWIN Transformer model, known for its ability to capture intricate image patterns, potentially leading to more accurate disease detection. To enhance transparency for farmers and agricultural experts, SwinLeafExplain integrates explainability features using different class activation maps, allowing them to visualize the model’s reasoning behind classifications. SwinLeafExplain achieved exceptional performance with an accuracy of 99.72%. Our proposed method significantly outperforms existing state-of-the-art methods in both accuracy and interpretability, providing valuable insights to aid agricultural decision-making.