Enhanced Transformer Framework for Multi-label Fine-grained Apple Leaf Disease Diagnosis

Apple leaf diseases can seriously affect apple yield and quality. This study presents a transformer-based fine-grained multi-label framework for the efficient classification of apple leaf diseases. The traditional methods that rely on specialized personnel face inefficiencies in large orchards. The framework achieved notable F1 scores, such as 0.855 for alternaria and 0.903 for brown spot, while significantly reducing computational complexity compared to conventional techniques. This novel approach not only advances disease classification but also provides crucial technical support for phenotyping platforms, aiding precision drug delivery and disease-resistant cultivar selection in apple cultivation.