taki0112 / SPADE-Tensorflow

Simple Tensorflow implementation of "Semantic Image Synthesis with Spatially-Adaptive Normalization" a.k.a. GauGAN, SPADE (CVPR 2019 Oral)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Suggestion: add requirements.txt

GzuPark opened this issue · comments

tqdm
numpy
pillow
scipy
opencv-python
tensorflow-gpu
keras

scipy==1.3.0 has a problem like below:
AttributeError: module 'scipy.misc' has no attribute 'imsave'

  • solution 1:
    imsave is deprecated! imsave is deprecated in SciPy 1.0.0, and will be removed in 1.2.0. Use imageio.imwrite instead.

  • solution 2:
    pip install scipy==1.2.0

this worked for me pip install scipy==1.0.0 also for requirements, if you use google colab most of these come preinstalled with free gpu, would suggest to add a colab notebook to this repo

@ak9250 Thank you, I know about colab :)
However, if I want to train by myself it will be interrupted due to their policy: session can alive 12h ~ 24h. And this repo needs approximately 4 ~ 5days for training. (EDIT) based on single V100

And, from the scipy documentation imsave can use until version 1.2.0 not only 1.2.0 also 1.0.0

(I'm training this great model with a provided dataset and single V100 like below:)
Screen Shot 2019-06-05 at 9 45 12 AM

Thank you for your suggestion.
I added to readme.

@GzuPark
Author changed the number of discriminator layer.
Therefore, I also changed. SPADE.py, main.py

so I recommend you to do the experiment again.

@taki0112
Thank you, I will do it asap