You can download the images for TIFA and AIGCIQA2023 datasets here.
Once downloaded, you can unzip them and inside the data-transfer
folder you should see two files: tifa_annotated_images.zip
and allimg.zip
. Unzip these two files, and you should get two folders: annotated_images
and allimg
.
The preference ranking data for TIFA can be accessed here: csvs/tifa.csv
. To setup the images, ensure there is a folder called annotated_images
in the same directory as the csv file.
The preference ranking data for AIGCIQA2023 can be accessed here: csvs/aigciqa2023.csv
. To setup the images, ensure there is a folder called allimg
in the same directory as the csv file.
The preference ranking for ECCV-Caption can be accessed here: eccv_caption.csv
. To setup the images, download the coco2014 val split from here. Unzip and then store the folder val2014
in the same directory as the csv file.
Download the images from here (originally from here). Extract the images and move the folder text2image_evaluation/img
is inside the same directory as the csv file.