AIGC related image dataset compilation

image datasets compilation

real image dataset:

  • laion-aesthetics containing the 600Mil image dataset used by stable diffusion: aesthetics scores >=5
    (download parquet files in data folder and run the following command to start downloading:)
    img2dataset --url_list data --input_format "parquet" --url_col "URL" --caption_col "TEXT" --output_format files --output_folder laion_images --processes_count 1 --thread_count 64 --resize_mode no
    

    official img2dataset command examples

stable diffusion generated image dataset:

midjourney generated image dataset: