CUB-200-2011
A frozen DINOv2 ViT-S/14 backbone with a dictionary of 1,024 concepts, up to 16 per image, and 16 quantized attributes of 64 values each. Across its 5,994 training images, a typical image has 16.
Red bellied Woodpecker · Downy Woodpecker · Red cockaded Woodpecker
Gadwall · Mallard · Brown Pelican
Long tailed Jaeger · Pomarine Jaeger · Black footed Albatross
Common Raven · American Crow · Brewer Blackbird
Wilson Warbler · Hooded Warbler · Kentucky Warbler
Glaucous winged Gull · Western Gull · Heermann Gull
House Sparrow · Tree Sparrow · Henslow Sparrow
Green Violetear · Anna Hummingbird · Ruby throated Hummingbird