Just looking at the Google results I find:
This challenge will target fine-grained classification, i.e. classification among categories which are are both visually and semantically very similar. This is a very difficult regime which is even challenging for humans without careful training, and is critical for establishing a more detailed understanding of the visual world.
First, being a fine-grained categorization problem,
there is little inter-class variation. For example the basset
hound and bloodhound share very similar facial characteristics
but differ significantly in their color, while the
Japanese spaniel and papillion share very similar color but
greatly differ in their facial characteristics.
So the upshot is that "fine-grained" means that the classes that are trying to be distinguished between are visually very similar.
In terms of a single image, it means that it is hard to see the differences across the entire image. In that case, your top image is fine-grained. The dog and bird images are not. Though if you zoomed into the hair of the dog or the feathers of the bird, I expect they would be.