"PlantDoc-Dataset-master"

"An external cross-dataset test set was constructed to further evaluate the adaptability of the proposed model under unseen image sources and environmental conditions. This test set comprises 50 plant disease images collected from publicly available sources. The selected disease categories align with representative categories found in the PlantDoc dataset, including late blight, powdery mildew, and others. These images cover a variety of visual conditions, such as different lighting environments, complex backgrounds, lesions of different sizes, and visually blurred disease symptoms. This test set is utilized exclusively for cross-dataset evaluation and is strictly excluded from the model training or fine-tuning processes. This configuration facilitates the assessment of whether a model trained on PlantDoc can maintain robust detection performance when applied to external plant disease images exhibiting distinct visual distributions."