xichenpan/ARLDM

Char-F1 and F-Acc score

Opened this issue · 1 comments

Previous works (VLCStoryGan, StoryDALL-E) also use Char-F1 and F-Acc scores to evaluate the character information of generated images. Have you evaluated these two indicators and what are the results?

@fengzhangyin Hi, we do not eval these two metrics. You can refer to the human eval results regarding relevant to see if our generated stories containing the correct character.