Generate images with specified noise seed

JiangguoZhang commented 1 year ago

In the GAN visualizer, I can change the latent or noise seed. However, in the gen_images.py, I can only specify the noise mode instead of the noise seed. Is there any way I can generate images in the command line with both specified latent and noise seeds?

jasuriy commented 4 months ago

@JiangguoZhang did you fix the issue? were you able to train the model with your own dataset?

JiangguoZhang commented 4 months ago

@jasuriy, thank you for your insights!

I haven't resolved the issue yet.

I've successfully trained the model using my dataset, and I found that StyleGAN2 performs better than StyleGAN3 for my needs. I'm looking to generate images that share the same features but differ based on the noise. However, by calling G(z, label, noise_mode=noise_mode), I couldn't find an option to specify the noise seed directly in the main model settings. The available noise_mode options are "const", "random", and "none". When set to "const", it seems I can't adjust the noise seed, though I noticed this functionality is present in visualizer.py. Could you guide me on how to utilize this feature or where I might be missing something?

jasuriy commented 4 months ago

Hi thank you for reply I really regret that right now I can't guide you on your issue because I am a beginner in this field and I would appreciate if you could guide me. I am having a very hard time to setup and train the model :( Could you please share the code you've used for training your dataset. thank you in advance Hope to hear from you soon.

Sincerely Jasurbek

[image: Mailtrack] https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality11& Sender notified by Mailtrack https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality11& 05/13/24, 01:25:40 PM

On Sat, May 11, 2024 at 7:13 AM Jiangguo Zhang @.***> wrote:

@jasuriy https://github.com/jasuriy, thank you for your insights!

I haven't resolved the issue yet.

I've successfully trained the model using my dataset, and I found that StyleGAN2 performs better than StyleGAN3 for my needs. I'm looking to generate images that share the same features but differ based on the noise. However, by calling G(z, label, noise_mode=noise_mode), I couldn't find an option to specify the noise seed directly in the main model settings. The available noise_mode options are "const", "random", and "none". When set to "const", it seems I can't adjust the noise seed, though I noticed this functionality is present in visualizer.py. Could you guide me on how to utilize this feature or where I might be missing something?

— Reply to this email directly, view it on GitHub https://github.com/NVlabs/stylegan3/issues/227#issuecomment-2105333288, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQKLHTRXFTWGQ5VTFQFOF23ZBVBBXAVCNFSM6AAAAAAWZMBET2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMBVGMZTGMRYHA . You are receiving this because you were mentioned.Message ID: @.***>

jasuriy commented 4 months ago

Sorry when I asked you the code you used, I meant the code of Stylegan2

[image: Mailtrack] https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality11& Sender notified by Mailtrack https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality11& 05/13/24, 09:23:42 PM

On Mon, May 13, 2024 at 1:29 PM Jasurbek Gopirjonov < @.***> wrote:

Hi thank you for reply I really regret that right now I can't guide you on your issue because I am a beginner in this field and I would appreciate if you could guide me. I am having a very hard time to setup and train the model :( Could you please share the code you've used for training your dataset. thank you in advance Hope to hear from you soon.

Sincerely Jasurbek

[image: Mailtrack] https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality11& Sender notified by Mailtrack https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality11& 05/13/24, 01:25:40 PM

On Sat, May 11, 2024 at 7:13 AM Jiangguo Zhang @.***> wrote:

@jasuriy https://github.com/jasuriy, thank you for your insights!

I haven't resolved the issue yet.

I've successfully trained the model using my dataset, and I found that StyleGAN2 performs better than StyleGAN3 for my needs. I'm looking to generate images that share the same features but differ based on the noise. However, by calling G(z, label, noise_mode=noise_mode), I couldn't find an option to specify the noise seed directly in the main model settings. The available noise_mode options are "const", "random", and "none". When set to "const", it seems I can't adjust the noise seed, though I noticed this functionality is present in visualizer.py. Could you guide me on how to utilize this feature or where I might be missing something?

— Reply to this email directly, view it on GitHub https://github.com/NVlabs/stylegan3/issues/227#issuecomment-2105333288, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQKLHTRXFTWGQ5VTFQFOF23ZBVBBXAVCNFSM6AAAAAAWZMBET2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMBVGMZTGMRYHA . You are receiving this because you were mentioned.Message ID: @.***>

NVlabs / stylegan3

Generate images with specified noise seed #227