Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

which weight file can be used? #5

Open
909041282 opened this issue Sep 25, 2018 · 15 comments
Open

which weight file can be used? #5

909041282 opened this issue Sep 25, 2018 · 15 comments

Comments

@909041282
Copy link

I have downloaded all file from google drive,but i don't know which can use to run your demo.
i have try the file in "weights_file/discriminator/epoch0480_loss0.560.hdf5",but the result is terrible when i use my photo as the input image.

@yh-iro
Copy link
Owner

yh-iro commented Sep 27, 2018

As noted in README.md, generalization is not good as author's. I tried some people not included in train images. As a result, some are so so (but with worse resolution) and some are collapsed. To achieve the quality like author's, I think it is required to put more technique into training, but still I can't make it clear. If I have more trial, maybe I will try smaller value for pixel loss coefficient and larger value for adversarial loss coefficient with time planning.

@KKKlose
Copy link

KKKlose commented Dec 4, 2018

I try the model 'epoch0480_loss0.560.hdf5',but i don't know where is the result img, could you tell me how to find it ?

@KKKlose
Copy link

KKKlose commented Dec 4, 2018

1543896802 1

@yh-iro
Copy link
Owner

yh-iro commented Dec 4, 2018

Please see misc/demo.py

I try the model 'epoch0480_loss0.560.hdf5',but i don't know where is the result img, could you tell me how to find it ?

@tema0000
Copy link

tema0000 commented Feb 3, 2019

I have been using 'demo.py' for a long time, but I still don't know where is the result front img. Could you help me how to find it? thank you.

01

@yh-iro
Copy link
Owner

yh-iro commented Feb 4, 2019

please see demo.py line 201.
Result image is stored in variable "pred_face". Maybe you have to save it by yourself.

I have been using 'demo.py' for a long time, but I still don't know where is the result front img. Could you help me how to find it? thank you.

01

@tema0000
Copy link

tema0000 commented Feb 4, 2019

Thanks for your reply. I have tested some images in the demo, but the results are not good enough. Is there a way to see the same direction in the result image eyes as the original image?

@jolinlinlin
Copy link

@tema0000 hello,I run demo.py but I can't get the result image from pred_faces,Can you tell me how to save result image?Thank you very much!

@lyupei
Copy link

lyupei commented Aug 8, 2019

@tema0000 hello,I run demo.py but I can't get the result image from pred_faces,Can you tell me how to save result image?Thank you very much!

you can save the result using cv2.imwrite('xxx.jpg', pred_faces[0,])

@hegc
Copy link

hegc commented Aug 28, 2019

the results are bad!

@DamDam123
Copy link

the results are bad!
Hello,I have tested some images in the demo,but I am having a problem now.
The problem is "TypeError: load_weights() missing 1 required positional argument: 'filepath'", Can you tell me how to solve this problem?Thanks you very much!

@sosoIrving
Copy link

the results are bad!
Hello,I have tested some images in the demo,but I am having a problem now.
The problem is "TypeError: load_weights() missing 1 required positional argument: 'filepath'", Can you tell me how to solve this problem?Thanks you very much!

the code in tpgan.py is filepath = filepath. You can delete " = filepath".
And you should and the weights at the corresponding place

@mumuyanyan
Copy link

the results of the projects are not good for my test pictures.

@ujwalakoriraj
Copy link

When I run demo.py, I only see 'input leye' in the console window, for a long time. I have given the test image path, the generator-weights path in demo.py, and the lcnn_extractor_weights path in the tpgan.py. Is there anything else I have to do or add? Should I press any button to get the output, because I see key == ord('s'), and key == ord('p') in the program?

@driver4567
Copy link

driver4567 commented Dec 7, 2023

The result I get for a sample image is really good, but for a personal image, the results are bad. Would that mean that the weights file needs more training (As I believe), or would I be doing something wrong?

The first is the output from images\x2.png
Output from x2

The second is an output from a personal image
Laura output

I thought maybe it might be the pitch (the angle of the pose) of the input face, but I tried some different inputs and got even worse results

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests