IIGROUP / MM-CelebA-HQ-Dataset

[CVPR 2021] Multi-Modal-CelebA: A Large-Scale Text-Driven Face Generation and Understanding Dataset

Home Page:https://arxiv.org/abs/2012.03308

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Upload 'caption.pickle' file

Kerry-zzx opened this issue · comments

commented

Dear author, thanks for share your code and dataset. Would you mind sharing the caption.pickle file of DFGAN and AttnGAN? Since I got a different result when using the pre-processing code of DFGAN to process the dataset. Thanks a lot,

commented

If you don't upload the 'caption.pickle' file, the pre-trained weight of DFGAN, DMGAN and AttnGAN may be not useful!!

Sorry for the delayed response. I am unable to provide the requested caption.pickle file. This file can be generated automatically during the training process. If you do not find the pre-trained models useful, you can easily train them yourself with the dataset, which is a straightforward process.

I just found the 'caption.pickle' file from #4.

This function loads captions.pickle from data_dir if it exists. If not, it will build such dictionary from scratch. You can use this pickle file I shared and put it in data_dir.