iglict / deepfacedrawing-jittor Goto Github PK

When the pytorch version can be released? I am interested in the manifold part of this work, and looking forward to the code release.

Thank you for your impressive work and generous sharing.

I have made it to generate some 2D images from the human sketches of my dataset with your code and pretrained models. However, I noticed that the generated 2D images all seem to be drawn from the front (canonical) view, while the viewpoint information, i.e. different observation angles, of the input skethes seem to be lost.

So I wonder if your model could keep the viewpoint information of the input sketches, and if it is possible how could I modify the code to keep the viewpoint information in the output 2D images. Thank you a lot.

Thanks for the great job. I want to know where can I download the training dataset used in the paper. I am looking forward to your reply.

Thanks for your great job and looking forward to the torch version.

Do you have any idea why there is a segfault when running the demo. THX :)

the Jitter version code is really cool, but it's not so general platform, Could author provide the PyTorch version, especially the training code?

Errors when run python3.7 demo.py Could you please give me some advice?
load_path ./Params/Combine/latest_net_DE_.pkl

Assertion failed:
Traceback (most recent call last):
  File "demo.py", line 13, in <module>
    ui = WindowUI()
  File "/home/zhaojiabei/DeepFaceDrawing-Jittor/WindowUI.py", line 43, in __init__
    self.input_scene = InputGraphicsScene(self.modes, self.brush_size,self.output_scene)
  File "/home/zhaojiabei/DeepFaceDrawing-Jittor/Input_mouse_event.py", line 105, in __init__
    self.predict_shadow()
  File "/home/zhaojiabei/DeepFaceDrawing-Jittor/Input_mouse_event.py", line 273, in predict_shadow
    sex=self.sex)
  File "/home/zhaojiabei/DeepFaceDrawing-Jittor/models/AE_Model.py", line 97, in get_inter
    generated_f = self.get_latent(input_image)
  File "/home/zhaojiabei/DeepFaceDrawing-Jittor/models/AE_Model.py", line 80, in get_latent
    mus_mouth = self.net_encoder(input_image)
  File "/home/zhaojiabei/anaconda3/envs/face_drawing/lib/python3.7/site-packages/jittor/__init__.py", line 511, in __call__
    return self.execute(*args, **kw)
  File "/home/zhaojiabei/DeepFaceDrawing-Jittor/models/networks.py", line 199, in execute
    mu = self.fc_mu(ten)
  File "/home/zhaojiabei/anaconda3/envs/face_drawing/lib/python3.7/site-packages/jittor/__init__.py", line 511, in __call__
    return self.execute(*args, **kw)
  File "/home/zhaojiabei/anaconda3/envs/face_drawing/lib/python3.7/site-packages/jittor/nn.py", line 1303, in execute
    x = layer(x)
  File "/home/zhaojiabei/anaconda3/envs/face_drawing/lib/python3.7/site-packages/jittor/__init__.py", line 511, in __call__
    return self.execute(*args, **kw)
  File "/home/zhaojiabei/anaconda3/envs/face_drawing/lib/python3.7/site-packages/jittor/nn.py", line 378, in execute
    x = matmul_transpose(x, self.weight)
  File "/home/zhaojiabei/anaconda3/envs/face_drawing/lib/python3.7/site-packages/jittor/nn.py", line 30, in matmul_transpose
    assert a.shape[-1] == b.shape[-1]
AssertionError

Hi,

Pretrained model produces random noise as a result. Anybody faced the same problem?

Hi there.
I was trying to run the demo. A got this nice error:

[e 0913 21:22:14.557574 60 executor.cc:442] 
=== display_memory_info ===
 total_cpu_ram: 7.658GB total_cuda_ram: 3.945GB
 hold_vars: 1639 lived_vars: 3180 lived_ops: 7531
 update queue: 1630/1630
 name: sfrl is_cuda: 1 used: 1.254GB(86.9%) unused: 194.2MB(13.1%) total: 1.443GB
 name: sfrl is_cuda: 1 used: 2.116GB(97.4%) unused: 56.98MB(2.56%) total: 2.172GB
 name: sfrl is_cuda: 0 used: 2.116GB(97.4%) unused: 56.98MB(2.56%) total: 2.172GB
 name: sfrl is_cuda: 0 used:   271KB(26.5%) unused:   753KB(73.5%) total:     1MB
 cpu&gpu: 5.788GB gpu: 3.615GB cpu: 2.173GB
===========================

[e 0913 21:22:14.557734 60 executor.cc:446] [Error] source file location: /home/jesus/.cache/jittor/default/g++/jit/_opkey0:broadcast_to_Tx:float32__DIM=7__BCAST=19__JIT:1__JIT_cuda:1__index_t:int32___opkey...hash:719812a59219cddf_op.cc

Apparently i ran out of RAM for this. Which is not unexpected I only have a measly 1050 at 4GB of RAM. But that got me wondering. What is your recommendation for GPU to run this? I want to know before i start looking for new video cards.

Can anyone share a video that how i can run this project?

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 410.104 Driver Version: 410.104 CUDA Version: 10.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla P100-PCIE... Off | 00000000:00:0D.0 Off | 0 |
| N/A 39C P0 43W / 250W | 16279MiB / 16280MiB | 0% Default

Is it necessary to increase the memory and how much configuration do you use?

It's really nice work you guys have done!
As the PyTorch version is mentioned as ""PyTorch version coming soon”， So when could we got this PyTorch version to study?

Currently I cannot find any good enough face sketch dataset on website. It would be helpful for many researchers if you give the source.

subj

Can you upload the dataset which you used to train your model?
Thanks

Hi,
First of all, I would like to thanks for your effort to provide us the source code and weight.
Actually, I can run it on my PC.
I have read your paper carefully, but I still concern about something, so I create this issue, and I hope anyone can help me to explain it.
1: I params folder, I see some file like man__feature.bin, female__feature.bin . But in your paper I can't map it with any network (CE, FM or IS). And i don't see you mention about sexuality in your paper. Can you help me to explain it.
2: Could you provide source code of net discriminator and file weight of it
Thanks and best regard

Hi !! first thanks for sharing your work, it's impressing, results are really good !!

I was wondering if it would be possible to share trained models (the Param folder) on another place than Baidu because it's nearly impossible to register on baidu without having a chinese phone number ?

thanks a lot !

This part of code was intent to perform Nearest Neighbor on feature_list, but it takes 4minutes on my PC.
However I found it works well on your online demo website. Is there ANY SOLUTIONS about that ?

I use flask for deployment, and the infer time is very slow, one second at a time

非常感谢！

i know it's because i run it on windows but please can any one tell how i switched to windows ?????

how to change the background?

got the below error while running:
on3.8 -I/home/karthik/.local/lib/python3.8/site-packages/pybind11/include
[i 0524 10:49:50.172849 52 compiler.py:898] extension_suffix: .cpython-38-x86_64-linux-gnu.so
[i 0524 10:49:50.427291 52 init.py:168] Total mem: 10.62GB, using 3 procs for compiling.
[i 0524 10:49:50.993130 52 jit_compiler.cc:20] Load cc_path: /usr/bin/g++
[i 0524 10:49:51.052464 52 init.py:249] Found mpicc(4.0.3) at /usr/bin/mpicc.
[i 0524 10:49:51.124027 52 compiler.py:653] handle pyjt_include/home/karthik/.local/lib/python3.8/site-packages/jittor/extern/mpi/inc/mpi_warper.h
Traceback (most recent call last):
File "demo.py", line 4, in
from WindowUI import WindowUI
File "/home/karthik/Documents/deeplearning/DeepFaceDrawing-Jittor/WindowUI.py", line 18, in
jt.flags.use_cuda = 1
RuntimeError: Wrong inputs arguments, Please refer to examples(help(jt.flags)).

Types of your inputs are:
self = flags,
arg = int,

The function declarations are:
void _set_use_cuda(int v)
void _set_use_cuda(bool v)

Failed reason:[f 0524 10:49:51.790183 52 cuda_flags.cc:30] Check failed: value==0 No CUDA found.

Could you release your complete code?

Thanks for your great jobs!
Can you release the code or methods to get the heat pictures?
And what is the useage of fake-mask.jpg in /heat?

Can this code be run on Google Colab?