envision-research / homointerpgan Goto Github PK

Hello! I am very interested in your work, especially the interpolation between dogs and cats. I saw this result at supplementary material. I want to interpolate without attribute, because there is no label or attribute in my dataset. So could you please tell me how to do?

I tried
python run.py attribute_manipulation -mp checkpoints/CelebA -sp checkpoints/CelebA/test/Smiling --filter_target_attr NOTSmiling -s 1 --branch_idx 0 --n_ref 5 -bs 8 --test_folder examples/aligned

and am getting weird results that lack identity preservation which is shown in the paper

vs

First of all, congratulation for this outstanding work, but I still have two small questions:

• Why we need "Rigorous Training" for attribute classifier ? Can you give a more exactly example?
• Can I consider T^k in interpolator as a "fliter" used to get rid of variables which are not related to correspond attribute from "feature code" F ?

Thanks!

Hi～ @yingcong .I encountered an error :"AttributeError: 'Namespace' object has no attribute 'random_crop_bias'" when I try to test the pretrained model which is downloaded from your link.Could you help me to solve the problem? And should I train the model 10000 epoches? Thank you!

Hi,

I would like to thank you first for your amazing work !

As the title said it, I would like to know if is there any way to generate an image without reference image at all ?
For example, generate a Woman face with only a Man face (without a Man face as reference).

test_selected_curve & attribute_manipulation need both as well a reference image to generate the interpolation (correct me if I''m wrong)

Thank you in advance !

Hi～ @yingcong Could you share the file of Rafd_image.csv or give me some suggestion about creating it?I'm confused of it.Thank you.

Thanks for sharing the nice work!

I have a few questions about the results of testing RaFD model on wild images:

1. As the images in RaFD dataset are well aligned, I am wondering whether you take any alignment steps there? Or just simply crop the face -> transform the expression -> place it back to the original image?
2. May I know whether you use all the images in RaFD to train the model or only the images under frontal camera (90 degree)?

Looking forward to your reply. Thanks.

I encountered this error when I tried to execute the training script. The line of code that caused the exception is seen here url. I am looking forward to your help.

Below is the detailed error message：

/home/shhs/anaconda3/envs/torch_1_0_py3_6/bin/python /home.bak/shhs/soft/pycharm-2019.1.1/helpers/pydev/pydevd.py --multiproc --qt-support=auto --client 127.0.0.1 --port 46783 --file /media/shhs/Peterou2/user/code/HomoInterpGAN/run.py attribute_manipulation -bs 8 -gpu 0
pydev debugger: process 6843 is connecting

Connected to pydev debugger (build 191.6605.12)
loading default VGG

• Total Images: 162770
  Traceback (most recent call last):
  File "/home.bak/shhs/soft/pycharm-2019.1.1/helpers/pydev/pydevd.py", line 1741, in
  main()
  File "/home.bak/shhs/soft/pycharm-2019.1.1/helpers/pydev/pydevd.py", line 1735, in main
  globals = debugger.run(setup['file'], None, None, is_module)
  File "/home.bak/shhs/soft/pycharm-2019.1.1/helpers/pydev/pydevd.py", line 1135, in run
  pydev_imports.execfile(file, globals, locals) # execute the script
  File "/home.bak/shhs/soft/pycharm-2019.1.1/helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile
  exec(compile(contents+"\n", file, 'exec'), glob, loc)
  File "/media/shhs/Peterou2/user/code/HomoInterpGAN/run.py", line 205, in
  engine.run()
  File "/media/shhs/Peterou2/user/code/HomoInterpGAN/run.py", line 200, in run
  exec ('self.{}()'.format(self.args.command))
  File "", line 1, in
  File "/media/shhs/Peterou2/user/code/HomoInterpGAN/run.py", line 161, in attribute_manipulation
  _, test_dataset = self.load_dataset()
  File "/media/shhs/Peterou2/user/code/HomoInterpGAN/run.py", line 106, in load_dataset
  csv_path='info/celeba-with-orientation.csv')
  File "/media/shhs/Peterou2/user/code/HomoInterpGAN/data/attributeDataset.py", line 220, in init
  f3 = self.frame.iloc[:, 1:] > 0
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 2108, in f
  res = self._combine_const(other, func)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/frame.py", line 5120, in _combine_const
  return ops.dispatch_to_series(self, other, func)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 1157, in dispatch_to_series
  new_data = expressions.evaluate(column_op, str_rep, left, right)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/computation/expressions.py", line 208, in evaluate
  return _evaluate(op, op_str, a, b, **eval_kwargs)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/computation/expressions.py", line 68, in _evaluate_standard
  return op(a, b)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 1128, in column_op
  for i in range(len(a.columns))}
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 1128, in
  for i in range(len(a.columns))}
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 1766, in wrapper
  res = na_op(values, other)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 1625, in na_op
  result = _comp_method_OBJECT_ARRAY(op, x, y)
  File "/home/shhs/anaconda3/envs/torch_1_0_py3_6/lib/python3.6/site-packages/pandas/core/ops.py", line 1603, in _comp_method_OBJECT_ARRAY
  result = libops.scalar_compare(x, y, op)
  File "pandas/_libs/ops.pyx", line 97, in pandas._libs.ops.scalar_compare
  TypeError: '>' not supported between instances of 'str' and 'int'

loading default VGG
/mnt/backup/project/ycchen/datasets/face/images/celeba_aligned/185954.jpg
Traceback (most recent call last):
File "run.py", line 192, in
engine.run()
File "run.py", line 187, in run
exec ('self.{}()'.format(self.args.command))
File "", line 1, in
File "run.py", line 156, in attribute_manipulation
img_out = [tmp for tmp in ref_loader]
File "/usr/local/lib/python2.7/dist-packages/torch/utils/data/dataloader.py", line 615, in next
batch = self.collate_fn([self.dataset[i] for i in indices])
File "/content/HomoInterpGAN/data/attributeDataset.py", line 163, in getitem
img = util.readRGB(self.files[index]).astype(np.float32)
File "/content/HomoInterpGAN/util/util.py", line 144, in readRGB
return img[:, :, [2, 1, 0]]
TypeError: 'NoneType' object has no attribute 'getitem'

at
python run.py attribute_manipulation -mp checkpoints/CelebA -sp checkpoints/CelebA/test/Smiling --filter_target_attr Smiling -s 1 --branch_idx 0 --n_ref 5 -bs 8 --test_folder examples/aligned

Very fancy idea!
When I was reading the paper, I could not find the supplementary material. Is it available?
Thanks~

why I meet this?