Table 3 Confusion matrix of the best ViT (ImageNet-21k TL) on the test set.

From: Non-invasive anemia detection from conjunctiva and sclera images using vision transformer with attention map explainability