This study presents a comprehensive investigation into the application of SegFormer, a transformer-based segmentation model, for the precise semantic segmentation of strawberry diseases. The research evaluates three distinct Mix Transformer encoders—MiT-B0, MiT-B3, and MiT-B5—to determine their effectiveness in disease detection and segmentation. The dataset consists of 2,450 raw images, expanded to 4,574 augmented images, and includes seven types of strawberry diseases: Leaf spot, Angular leaf spot, Anthracnose rot, Blossom blight, Gray mold, Powdery mildew on fruit, and Powdery mildew on leaves. The study employs the Segment Anything Model (SAM) integrated into the Roboflow annotation tool for efficient dataset preparation. The results indicate that MiT-B3 and MiT-B5 outperform MiT-B0 in terms of segmentation accuracy and performance. MiT-B3 demonstrates rapid adaptation and consistent training and validation performance, while MiT-B5 provides efficient learning with occasional fluctuations, achieving the most precise segmentation in general. The study also highlights the importance of early stopping and model checkpointing in preventing overfitting and ensuring model stability. The findings suggest that MiT-B3 is particularly effective for real-world applications due to its rapid adaptation and reliable detection capabilities. The research contributes to the field of agricultural disease detection by providing insights into the selection of appropriate encoders for different disease detection tasks and demonstrating the potential of SegFormer for future research and interdisciplinary collaboration.This study presents a comprehensive investigation into the application of SegFormer, a transformer-based segmentation model, for the precise semantic segmentation of strawberry diseases. The research evaluates three distinct Mix Transformer encoders—MiT-B0, MiT-B3, and MiT-B5—to determine their effectiveness in disease detection and segmentation. The dataset consists of 2,450 raw images, expanded to 4,574 augmented images, and includes seven types of strawberry diseases: Leaf spot, Angular leaf spot, Anthracnose rot, Blossom blight, Gray mold, Powdery mildew on fruit, and Powdery mildew on leaves. The study employs the Segment Anything Model (SAM) integrated into the Roboflow annotation tool for efficient dataset preparation. The results indicate that MiT-B3 and MiT-B5 outperform MiT-B0 in terms of segmentation accuracy and performance. MiT-B3 demonstrates rapid adaptation and consistent training and validation performance, while MiT-B5 provides efficient learning with occasional fluctuations, achieving the most precise segmentation in general. The study also highlights the importance of early stopping and model checkpointing in preventing overfitting and ensuring model stability. The findings suggest that MiT-B3 is particularly effective for real-world applications due to its rapid adaptation and reliable detection capabilities. The research contributes to the field of agricultural disease detection by providing insights into the selection of appropriate encoders for different disease detection tasks and demonstrating the potential of SegFormer for future research and interdisciplinary collaboration.