Hi
seungyounshin / adaptive-wing-loss-for-robust-face-alignment-via-heatmap-regression Goto Github PK
View Code? Open in Web Editor NEWAdaptive-Wing-Loss-for-Robust-Face-Alignment-via-Heatmap-Regression
Adaptive-Wing-Loss-for-Robust-Face-Alignment-via-Heatmap-Regression
Hi
I want to make a lightweight landmark detection model like OpenSeeFace, but there is no training code in OpenSeeFace repo. Your repo has only HourglassNet, but can you make a lightweight model for mobile? Do you have any idea what is good if you make it?
When I placed 300W and ran train.py, I got the following error. Do you know the solution?
> python train.py 11:45:54
train start!
Traceback (most recent call last):
File "train.py", line 47, in <module>
out = model(img)
File "/home/anasys/.pyenv/versions/miniconda3-4.3.30/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in __call__
result = self.forward(*input, **kwargs)
File "/mnt/efs/personal/kuroyanagi/work/20200528-1/Adaptive-Wing-Loss-for-Robust-Face-Alignment-via-Heatmap-Regression/models/hourglass.py", line 179, in forward
y = self.coordconv[i](x,boundary_pred)
File "/home/anasys/.pyenv/versions/miniconda3-4.3.30/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in __call__
result = self.forward(*input, **kwargs)
File "/mnt/efs/personal/kuroyanagi/work/20200528-1/Adaptive-Wing-Loss-for-Robust-Face-Alignment-via-Heatmap-Regression/models/CoordConv.py", line 67, in forward
ret = self.addcoords(x, boundary_map)
File "/home/anasys/.pyenv/versions/miniconda3-4.3.30/lib/python3.6/site-packages/torch/nn/modules/module.py", line 493, in __call__
result = self.forward(*input, **kwargs)
File "/mnt/efs/personal/kuroyanagi/work/20200528-1/Adaptive-Wing-Loss-for-Robust-Face-Alignment-via-Heatmap-Regression/models/CoordConv.py", line 45, in forward
xx_channel, zero_tensor)
RuntimeError: Expected tensor to have cpu DeviceType, but got tensor with cuda DeviceType (while checking arguments for CPU_tensor_apply)
>
Would you please update the evaluate result on 300W full
hi, thanks for your work. I am very interested in it.
I have a question about cropping the image in trainset.
In face300w.py, line 63, when cropping the image according to the coordinates of landmarks, why use "miny-face_h" for the coordinates of the upper left corner, and use "+ or - crop_pad" for all others?
Hope for your reply.
hi~I have learned your code and gained a lot. And I have two questions.
The first one is about the learning rate. The AWing paper mentioned that the initial learning rate is 1e-4, and you set it to 1e-6 in the code. Is the learning rate of 1e-6 better?
The second one is about the function '_nms'. In eval.py, there are two lines of code, "from utils.post_process import _nms out=_nms(out)". There is no "post_process" file here, what is the function of โ_nms"? Will the lack of it affect the output of the result?
Hope for your reply.
The function '_nms'. In eval.py, there are two lines of code, "from utils.post_process import _nms out=_nms(out)". There is no "post_process" file here, what is the function of โ_nms"? Will the lack of it affect the output of the result?
Hope for your reply.
ret = torch.cat([ret, xx_channel, yy_channel], dim=1)
should be
ret = torch.cat([ret, xx_boundary_channel, yy_boundary_channel], dim=1)
?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.