hrlty / tp-gan Goto Github PK

Hello,Can you provide the pred weight? And I get some cropped dataset ,but I don't have label files. How to get label file from ".5pt"file?

I would like to run the code and try to reproduce the results in your paper. How can I do that?
Could you please describe the steps to train the network?

Hi, Thanks for sharing the codes! Could you please tell us the structure of the dataset

main folder and subfolders to make another simple dataset to train this model since I can not bay CMU Multi-PIE Face Database

def GetFeatureParts(self, img_resize, filename, label=False):
#crop four parts
trans_points = np.empty([5,2],dtype=np.int32)
if True:#not label:
featpath = self.feat5ptDir + filename[0:-15] + '_trans.5pt'

Can you please also share pretrained weights for DeepFace and the whole model?

Hi,
Thank you for your work.
As mention is the title, is there any guide to do the inference?

Thank you,
tsly

I managed to start training on the sample data. However after a couple of batches all the lossess instantly turn to nan and the saved output images become black. I'm using python3 and tensorflow 1.4.1. Any ideas?

Has anyone run the code successfully? Does anybody have a trained model , thanks very much

Thank you for sharing your code. I want to run a test on the model, is it possible to provide a pretrained model as well as the operating steps? Thank you

Hi, Thanks for offering this work.
I'm curious how many identities used in the training set. I notice that it's 340 in the function "FeaturePredict". Waiting for reply.
Thanks a lot.

Can you share the trained model file, I want use it to test face synthetize?
I have not enough data for training TP-GAN model.
Thank you!

I tested the shared model(DeepFace168.pickle) following the setting in paper(cvpr2015, light-cnn). But the result was not good. I obtained the following result:
15°: 56.7%
30°: 34.0%
45°: 14.7%
60°: 4.9%
Some errors may exist in my program. But I wonder if someone has tested the DeepFace168.pickle on Multi-pie Dataset?

Can you share your file by https://pan.baidu.com? It is difficult to vist drive.google.com now!

How do you perform face alignment?
Is it to use affine transformation after detecting the key points of the face?

thank u!

Hello!

I'm here also really interested in having pre-trained weights for this (as I can't access dataset to train my own model)

Thanks in advance!

Could you give a A detail README file and some notes in code?

This initial version is too raw. As a fish, I can not run it.

Traceback (most recent call last):
File "TP_GAN-Mar6FS.py", line 1132, in
tf.app.run()
File "/home/anaconda2/lib/python2.7/site-packages/tensorflow/python/platform/app.py", line 48, in run
_sys.exit(main(_sys.argv[:1] + flags_passthrough))
File "TP_GAN-Mar6FS.py", line 1128, in main
sample_dir=FLAGS.sample_dir)
File "TP_GAN-Mar6FS.py", line 127, in init
self.loadDeepFace(self.DeepFacePath)
File "TP_GAN-Mar6FS.py", line 866, in loadDeepFace
logging.error(("File '%s' not found. "), DeepFacePath)

Hello,how to test face images?can you provide a pre-model?

I encounter the following long error running the TP_GAN-Mar6FS.py program:

.............

Traceback (most recent call last):

File "D:\python\projects\reidentification\face_frontalization\frontalization-master\4_TP-GAN-master\TP_GAN-Mar6FS.py", line 1059, in
tf.compat.v1.app.run()

File "C:\Users\rajak\anaconda3\envs\deep_learning\lib\site-packages\tensorflow_core\python\platform\app.py", line 40, in run
_run(main=main, argv=argv, flags_parser=_parse_flags_tolerate_undef)

File "C:\Users\rajak\anaconda3\envs\deep_learning\lib\site-packages\absl\app.py", line 295, in run
flags_parser,

File "C:\Users\rajak\anaconda3\envs\deep_learning\lib\site-packages\absl\app.py", line 364, in _run_init
flags_parser=flags_parser,

File "C:\Users\rajak\anaconda3\envs\deep_learning\lib\site-packages\absl\app.py", line 217, in _register_and_parse_flags_with_usage
if FLAGS.only_check_args:

File "C:\Users\rajak\anaconda3\envs\deep_learning\lib\site-packages\absl\flags_flagvalues.py", line 474, in getattr
raise AttributeError(name)

AttributeError: only_check_args
.............

您好，您用的pickle模块的版本号是多少？
我在执行这一句时
self.data_dict = pickle.load(file)
遇到错误
·UnicodeDecodeError: 'gbk' codec can't decode byte 0x80 in position 0: illegal multibyte sequence·

I'm running your code, and I got the warning in log.
Where can I get evaluation_rank.sh?

hello, i can not find this file. There just exist DeepFace168.pickle. Can you answer my question? Thank you very much.

Is anyone run the model sucessfully? I use the CAS-PEAL-R1 dataset to train the model but the loss is nan. Maybe my data-preprocessing is not correct. Anyone can help me?

hello,thank you for sharing your code. I am interesting in TP-GAN and want to run a test, can you provide train dataset or a pretrained modelfor me? Thank you