Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
A pretrained CNN backbone takes an image, represented by its pixel values, and creates a low-resolution feature map of it.