ViT-B-VTN (3 layers, ImageNet pretrain)

Reported on 2 benchmarks across 1 task · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.