search

yipoh / AesBench

An expert benchmark aiming to comprehensively evaluate the aesthetic perception capacities of MLLMs.
Apache License 2.0
215 stars 6 forks source link

can not find the expert-answered aesthetic interpretations in dataset #9

Open BetterZH opened 1 month ago

BetterZH commented 1 month ago

Hi, I have downloaded the EAPD_release.zip from Google Drive. But I noticed that the AesBench_evaluation.json file is incomplete, containing only the "Question" field in the AesI_data section, while the expert-answered aesthetic interpretations seem to be missing. If possible, could you kindly provide the complete dataset?

image
dylanqyuan commented 1 month ago

Thank you for your attention in AesBench 😄 At present, the data you mentioned above will NOT be made public. We have made this decision to ensure that these data are not used for model training.

BetterZH commented 1 month ago

Thank you for your attention in AesBench 😄 At present, the data you mentioned above will NOT be made public. We have made this decision to ensure that these data are not used for model training.

I fully understand your decision not to make the data public to avoid it being used for training. However, my goal is solely to test a model's aesthetic explanation capabilities for evaluation purposes.

If there's any way to access the data under these terms, I would greatly appreciate it and will fully comply with any restrictions to ensure proper use.

Thank you for considering my request.