test error about tbgan

[Prometheus1995]

Hello, i experienced the same Error Message. If anyone finds a solution i would be grateful :D

[Prometheus1995]

So i tried out different methods. I can create the tf-records for the mnist dataset but running them with train.py will fail with the same statement. I tried running it on google colab with gpu support maybe TBGAN is having trouble with that.

[Prometheus1995]

So after some more tryouts i get this Error executing the test.py with the Model

[barisgecer]

Hi guys,
The first error seems to be weird. tfutil.py has only comment line at line 236:

TBGAN/tfutil.py

Line 236 in 6c7134e

#----------------------------------------------------------------------------

You could check if you have the generate_interpolation_images function in util_scripts.py.

Finally, please re-clone the whole repository from scratch. I have played with the branches lately that might have confused your local git.

Hope that helps. Let me know.
Baris

[Prometheus1995]

Hello, i started from scratch in colab 5 days ago and redid the same 2 days ago nothing changed since then. Looked for the generate_interpolation_images function in util_scripts and i still wondered if it is some problem with the .mp4 in the image generation? But if its working for you it may be a problem with the python path finding. Gonna check that tomorrow :D and thanks for the help :)

[barisgecer]

Please ignore mp4 codes in generate_interpolation_images. They are inherited from generate_interpolation_video but they are not actually used. I believe it is regarding path finding because obviously, it cannot import generate_interpolation_images function.

Also, have you checked the requirements? The code is based on Progressive Growing GAN from NVIDIA, so you can check if you can successfully run their code first. https://github.com/tkarras/progressive_growing_of_gans

[flynnamy]

Hi,how to solve like this? @barisgecer

[barisgecer]

@flynnamy I couldn't reproduce it. Did you try running https://github.com/tkarras/progressive_growing_of_gans ? Our code is built based on Pro-GAN. If you can run it, you should be able to run ours as well.

[XuZhengzhuo]

I have the same problem ...

[aashishrai3799]

I was able to solve this issue by installing menpo3d (pip install menpo3d).

[XuZhengzhuo]

I have the same problem ...

I directly replace the following code in test.py:

    tfutil.call_func_by_name(**config_test.train)

as:

    from util_scripts import generate_interpolation_images
    config_test.num_gpus = 1
    config_test.desc = 'interpolation-images-' + str(config_test.train.run_id)
    generate_interpolation_images(run_id=32, grid_size=[1,1], duration_sec=5.0, smoothing_sec=0.1)

and it reminds me to install menpo3d. Then it works.

[barisgecer]

Thanks @aashishrai3799 and @XuZhengzhuo for letting us know what was the issue. I didn't realize that I already have menpo3d and the error is not self-explanatory anyway. I am glad we finally have a solution, I will add this to the readme file. Thanks again.

[XuZhengzhuo]

Thanks @aashishrai3799 and @XuZhengzhuo for letting us know what was the issue. I didn't realize that I already have menpo3d and the error is not self-explanatory anyway. I am glad we finally have a solution, I will add this to the readme file. Thanks again.

Hi, how can I generate a 3D shape from a signal image (not from a random latent code)？I mean is there something like a projector.py in stylegan to encode a image into a latent code in your repo? Many thanks!!!

[barisgecer]

Unfortunately, my method does not have this feature. Since it is based on ProGAN , it does not inherit features from StyleGAN. Nevertheless, I believe it is doable like done in GANFit.

[flynnamy]

I have the same problem ...

I directly replace the following code in test.py:

    tfutil.call_func_by_name(**config_test.train)

as:

    from util_scripts import generate_interpolation_images
    config_test.num_gpus = 1
    config_test.desc = 'interpolation-images-' + str(config_test.train.run_id)
    generate_interpolation_images(run_id=32, grid_size=[1,1], duration_sec=5.0, smoothing_sec=0.1)

and it reminds me to install menpo3d. Then it works.

I modify like this,but it has error,how to solve like this? Thanks! @XuZhengzhuo

[XuZhengzhuo]

I have the same problem ...

I directly replace the following code in test.py:

    tfutil.call_func_by_name(**config_test.train)

as:

    from util_scripts import generate_interpolation_images
    config_test.num_gpus = 1
    config_test.desc = 'interpolation-images-' + str(config_test.train.run_id)
    generate_interpolation_images(run_id=32, grid_size=[1,1], duration_sec=5.0, smoothing_sec=0.1)

and it reminds me to install menpo3d. Then it works.

I modify like this,but it has error,how to solve like this? Thanks! @XuZhengzhuo

You may need to email the authors first and sign an End User License Agreement. The authors will give you the pretrain result and place it into '/results' directory.