nile649 / POLY-GAN

Poly-GAN: Multi-Conditioned GAN for Fashion Synthesis. (Not updating).

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Steps to run the code

dubesar opened this issue · comments

Hey, there's some confusion regarding running the code. Can you create a copy of google colab for this code so that we can directly run on colab.

Hey @dubesar , I had to modify a couple of things so far to run the code:

  1. In order to run the test.py the parameter was --stage "Stage3", not --stage "Refine"
  2. The test.py, line 85 is using absolute path plt.imsave("{}/{}_{}_ref.jpg".format(save_ref,f,img1[:-6]),resize(plt.imread("/home/np9207/vton/data/{}/image/{}_0.jpg".format(opt.datamode,img1[:-6])),(128,128))), I changed it too.

I am trying to encapsulate all the models needed to generate the final result, given a cloth and a person image, although I am having trouble to use the pose estimator and segmentation models.

commented

Here is my steps to run this whole testing process through on Colab, the final result is not satisfying as I expected. Probably I'm doing it in a wrong way.

Model image:
1

cloth image:
2

final result:
3

  1. Get repo and data
!git clone https://github.com/nile649/POLY-GAN.git
%cd POLY-GAN
!cp -r /content/drive/MyDrive/Dataset/poly-gan/pre_trained_models/ ./
!cp -r ./pre_trained_models/data ./
!unzip ./data/data.zip
  1. show some data
from PIL import Image
im = Image.open("/content/POLY-GAN/data/test/image/000001_0.jpg")
im
  1. change code

in test.py, line 10, change

from skimage.filters import threshold_otsu,threshold_adaptive

to

from skimage.filters import threshold_otsu,threshold_local

in line 97, also change

binary = threshold_adaptive(temp2[:,:,0], block_size, offset=0)

to

binary = threshold_local(temp2[:,:,0], block_size, offset=0)

In line 85, change

"/home/np9207/vton/data/{}/image/{}_0.jpg"

to

"./data/{}/image/{}_0.jpg"

Note: if you don't want to change threshold_adaptive, you need to install scikit-image==0.14.2 as mentioned in requirements.txt, which may take you some minutes

  1. run

stage1

!python test.py --datamode test --stage Stage1 --model_image 000001_0.jpg --reference_image 000347_1.jpg

stage2

!python test.py --datamode test --stage Stage2 --model_image 000001_0.jpg --reference_image 000347_1.jpg

stage3

!pip install matplotlib==3.0.2
!!python test.py --datamode test --stage Stage3 --model_image 000001_0.jpg --reference_image 000347_1.jpg

Then you cannot find final result at results/test/Stage3. It's at ./results/test/Stage3temp_res.