There is a medical image dataset that has 3 labels, but the whole dataset is around 1600 images

The dataset seems small compared to the datasets in the course, and I am worried that the model may overfit.