This is a benchmark dataset for conditional image generation, containing 990 test examples. The benchmark contains a diverse set of guidance signals, including depth maps, bounding boxes, art styles, and more to produce a wide array of conditional image generation scenarios.
original_images/: Contains 990 original reference images
images/: Contains 1710 condition images
benchmark990v2.json: The metadata file of the dataset… See the full description on the dataset page:
https://huggingface.co/datasets/Dipsy0830/MultiRef.
