papulke / face-of-art Goto Github PK

Dear author,

Thanks for sharing your code!

I found some problems when testing the scripts. I first downloaded the datasets as well as the pre-trained models, and then changed the related file paths and executed the script directly python3 predict_landmarks.py

However, I first got some complaints about the convention from py2 to py3 such as using print() rather than print at menpofit/clm/algorithm/gd.py:363

And then I got the following messages:

Traceback (most recent call last):
File "predict_landmarks.py", line 58, in
reuse=reuse, map_to_input_size=map_landmarks_to_original_image)
File "/home/chloe/face-of-art/deep_heatmaps_model_fusion_net.py", line 822, in get_landmark_predictions
p_pdm_lms = feature_based_pdm_corr(lms_init=init_lms, models_dir=pdm_models_dir, train_type='basic')
File "/home/chloe/face-of-art/pdm_clm_functions.py", line 160, in feature_based_pdm_corr
pdm_temp = pickle.load(filehandler, fix_imports=True, encoding="latin1")
_pickle.UnpicklingError: unpickling stack underflow

I simply changed the encoding way from "latin1" to "bytes" as a solution, however, I am not sure if it is correct or not.

And then, I got this error:

Traceback (most recent call last):
File "predict_landmarks.py", line 58, in
reuse=reuse, map_to_input_size=map_landmarks_to_original_image)
File "/home/chloe/face-of-art/deep_heatmaps_model_fusion_net.py", line 822, in get_landmark_predictions
p_pdm_lms = feature_based_pdm_corr(lms_init=init_lms, models_dir=pdm_models_dir, train_type='basic')
File "/home/chloe/face-of-art/pdm_clm_functions.py", line 165, in feature_based_pdm_corr
part_lms_pdm = pdm_correct(lms_init[part_inds], pdm_temp)
File "/home/chloe/face-of-art/pdm_clm_functions.py", line 86, in pdm_correct
pdm_model.set_target(PointCloud(init_shape))
File "/home/chloe/.local/lib/python3.6/site-packages/menpo/base.py", line 229, in set_target
self._target_setter_with_verification(new_target) # trigger the update
File "/home/chloe/.local/lib/python3.6/site-packages/menpo/base.py", line 244, in _target_setter_with_verification
self._verify_target(new_target)
File "/home/chloe/.local/lib/python3.6/site-packages/menpo/base.py", line 266, in _verify_target
if self.target is None:
File "/home/chloe/face-of-art/menpofit/modelinstance.py", line 114, in target
return self._target
AttributeError: 'OrthoPDM' object has no attribute '_target'

Did I miss any settings? Or what should I do to run through this script?

Thank you for your time!

Great work! I would like to check what the license is of this code and model in this repo?

Thank you for sharing the code, this is exactly what I need! I must get facial landmarks on art portraits. But I'm struggling to get it working since a week.
I´m using Python=2.7, menpo=0.8.1. Installing compatible packages was already a challenge but now I´m getting this error when I try to run predict_landmarks.py::

NotFoundError (see above for traceback): Tensor name "heatmaps_network/conv_1/conv/bias" not found in checkpoint files ./model_foa/deep_heatmaps-60000.data-00000-of-00001
         [[Node: save/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save/Const_0_0, save/RestoreV2/tensor_names, save/RestoreV2/shape_and_slices)]]

and

tensorflow.python.framework.errors_impl.DataLossError: Unable to open table file ./model_foa/deep_heatmaps-60000.data-00000-of-00001: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
         [[Node: save/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save/Const_0_0, save/RestoreV2/tensor_names, save/RestoreV2/shape_and_slices)]]

I downloaded the model weights from this link and put it into the model_foa folder. I´m not even sure if this is correct?

However there is no deep_heatmaps-60000 and the downloaded files do not work. I get the above error when using the .meta, .index file or .data... file from dropbox.

I guess the correct file is missing? what is it supposed to be or where do I get it?

unable to create conda env with newer glibc.

any idea how to overcome this?

I am trying to open pdm_models. But when I try to open these files I am getting and error as below.

pdm_clm_functions.py", line 158, in feature_based_pdm_corr
    pdm_temp = pickle.load(filehandler)
ModuleNotFoundError: No module named 'copy_reg\r'

Any idea why I can not load these models. I tried using python 2 and 3 both.

Thank you in advance.

Thanks for this great work！

Can you tell me how to train with my own dataset？ I am a little confused with the construction of the dataset.

Can you help me? Thanks so much.

hi~ first of all, thanks a lot for your awesome work and open-source codes.

It seems that predict_landmarks.py only produces the landmarks.

How to obtain the transfer results in the teaser, namely the rendered faces given the reference art works?

try:
pdm_temp = pickle.load(open(temp_model, "rb"))
except UnicodeDecodeError:
pdm_temp = pickle.load(open(temp_model, "rb"), fix_imports=True, encoding="latin1")

2021-08-02 15:47:16.855329: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7
Traceback (most recent call last):
File "/students/julyedu_623899/face_of_art/face-of-art/predict_landmarks.py", line 61, in
reuse=reuse, map_to_input_size=map_landmarks_to_original_image)
File "/students/julyedu_623899/face_of_art/face-of-art/deep_heatmaps_model_fusion_net.py", line 828, in get_landmark_predictions
p_pdm_lms = feature_based_pdm_corr(lms_init=init_lms, models_dir=pdm_models_dir, train_type='basic')
File "/students/julyedu_623899/face_of_art/face-of-art/pdm_clm_functions.py", line 169, in feature_based_pdm_corr
pdm_temp = pickle.load(filehandler)
EOFError: Ran out of input

Hello, I have an error while running the file.
Hello, I am a novice in the field of face key detection, thank you very much for sharing the model, I need to use your key detection model to detect the key points of cartoon characters in the task, so I want to ask you about this error?

hi, when I run predict_landmarks.py
os: ubuntu 18
Traceback (most recent call last):
File "predict_landmarks.py", line 59, in
reuse=reuse, map_to_input_size=map_landmarks_to_original_image)
File "/mnt/g/github/face-of-art/deep_heatmaps_model_fusion_net.py", line 822, in get_landmark_predictions
p_pdm_lms = feature_based_pdm_corr(lms_init=init_lms, models_dir=pdm_models_dir, train_type='basic')
File "/mnt/g/github/face-of-art/pdm_clm_functions.py", line 159, in feature_based_pdm_corr
pdm_temp = pickle.load(filehandler)
ModuleNotFoundError: No module named 'copy_reg\r'

I'm on Mac M2. When creating the v_env I can only work with a version of python3.8 or above. this causes issues as I am unable to install tensor flow 1.14 which is the only one that works with this code?

I was not able to install the code with python 2.7 because of the version conflicts on menpo.
I think a requirements.txt file with version fixing would be extremely useful to help others running the codes. Also, sample data and folders for testing would help too.
It is a very fun project and I would love to be able to repeat it.