Datasets:
Do the subsets have common images?
In the subset "DiffusionDB 2M", there are parts 1 to 2000.
In the subset "DiffusionDB Large", there are parts 1 to 14000.
My question is, are the parts 1 to 2000 in subset "DiffusionDB 2M" the exact same as the parts 1 to 2000 in "DiffusionDB Large"?
Hello @Mushfi , sorry for the late reply! "DiffusionDB 2M" is not exactly in the parts 1 to 2000 in "DiffusionDB Large". Sorry for the inconvenience!
The key difference between these two subsets is the number of images per prompt. Many Stable Diffusion users generated multiple images using the same prompt + hyperparameters. In these cases, "DiffusionDB 2M" only randomly selects one image to include, while "DiffusionDB Large" collects all images. Therefore, images in "DiffusionDB 2M" are more spread out in "DiffusionDB Large".