Undesired generated results

[CiciBEASTARS]

Thanks for your wonderful work!!! I download the pre-trained models latest.pth and latents_eval_50k.pth from the Google drive link you provided, and run the command below: python main.py \ --jobname author_release --latents_eval latents_eval_50k.pth \ --exp_mode visualize_video --n_styles 11 --num_frames 250 \ --given_subject_list 1000-1010 --style_id 7

However, I found the generated source and stylized images really weird, the 1000_source.png is like this: and the stylized image with style_id = 1 :

Do u know where might the problem be? Am I missing something, or is there anything wrong with the pre-trained checkpoints? Really looking forward to your reply! Thanks in advance!

[junzhezhang]

Thank you for your interest in our work! I will look into it by early next month :)

[2680347338]

Same problem, is it solved now please?

[CiciBEASTARS]

Same problem, is it solved now please?

Haven't solved it yet.

[PeizhiYan]

Same problem occurred

[junzhezhang]

Hi all, I am sincerely sorry for the late update! I will only be able to work on it during the year-end holidays. I will update you guys here.

[PeizhiYan]

I tried to manually set the style_id in the main.py code (as shown in the following screenshot),

Following are example generated images I got:

style_id = 0

style_id = 1

style_id = 2

style_id = 3

style_id = 4

style_id = 5

[junzhezhang]

Hi All,

The bug has been fixed. Sorry for the late response.

Best, Junzhe

[heeeyk]

Hi All,

The bug has been fixed. Sorry for the late response.

Best, Junzhe

Thanks for your update! I have run the new code and the correct result appeared. However, for some styles, there are still some artifacts. For example, style_id 2 and 4, as shown below：

Is this normal?

[Xia-zx]

Thanks for your wonderful work!!! I download the pre-trained models latest.pth and latents_eval_50k.pth from the Google drive link you provided, and run the command below: python main.py \ --jobname author_release --latents_eval latents_eval_50k.pth \ --exp_mode visualize_video --n_styles 11 --num_frames 250 \ --given_subject_list 1000-1010 --style_id 7

However, I found the generated source and stylized images really weird, the 1000_source.png is like this: and the stylized image with style_id = 1 :

Do u know where might the problem be? Am I missing something, or is there anything wrong with the pre-trained checkpoints? Really looking forward to your reply! Thanks in advance!

Where can I download the pre-training model latest.pth and latents_eval_50k.pth provided by the author？