twobranchdracaena / openface-pytorch Goto Github PK
View Code? Open in Web Editor NEWPyTorch model of OpenFace
PyTorch model of OpenFace
Hi,
I download your repo, and run the test.py, but there is a bug that lead to crash!
it shows: RuntimeError: in-place operations can be only used on variables that don't share storage with any other variables, but detected that there are 2 bojects sharing it.
layer/normalize.py: line 28 in forward
x[I].div_(norm.data[I][0]
torch/autograd/variable.py,line321 in div_
return DivConstant.apply(self, other, True)
I have run the model against the lennon, eric_clapton test pics and the results are wrong.
The results should be https://cmusatyalab.github.io/openface/demo-2-comparison/
The model produces:
lennon-1 lennon-2 tensor(1.0710, device='cuda:0')
lennon-1 clapton-1 tensor(2.0752, device='cuda:0')
lennon-1 clapton-2 tensor(1.1153, device='cuda:0')
lennon-2 clapton-1 tensor(1.0742, device='cuda:0')
lennon-2 clapton-2 tensor(1.8671, device='cuda:0')
clapton-1 clapton-2 tensor(2.1628, device='cuda:0')
The pics that I have used for lennon2 and clapton2 are not exactly the same as in the web but anyway the results with both set of pics is not below 0.99 for the real matches.
Got this error while running test.py
In the README, it says:
The conversion is mostly done by clarwin's convert_torch_to_pytorch, with some added layers, e.g. Inception.
Please can you explain what modifications are necessary to the convert_torch_to_pytorch script, or publish your patches, so that it is possible for others to convert their own custom FaceNet .t7 models into PyTorch format?
In the convert_torch_to_pytorch repo, they have explicitly stated that FaceNet models are not supported, because of the Inception layers:
clcarwin/convert_torch_to_pytorch#2
clcarwin/convert_torch_to_pytorch#5
It's great that you have found a way to get it to work. Please can you share how you did it? This is useful to know for those of us that want to convert our own custom FaceNet .t7 model rather than use nn4.small2.v1.t7
Thanks very much.
Running the accompanying test.py I get the following error:
RuntimeError Traceback (most recent call last)
<ipython-input-10-a0eb75f7e274> in <module>()
11
12 # CPU
---> 13 output = model(Variable(input))
14 print('CPU done')
15
~/envs/face-recognition/lib/python3.5/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
323 for hook in self._forward_pre_hooks.values():
324 hook(self, input)
--> 325 result = self.forward(*input, **kwargs)
326 for hook in self._forward_hooks.values():
327 hook_result = hook(self, input, result)
~/envs/face-recognition/lib/python3.5/site-packages/torch/nn/modules/container.py in forward(self, input)
65 def forward(self, input):
66 for module in self._modules.values():
---> 67 input = module(input)
68 return input
69
~/envs/face-recognition/lib/python3.5/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
323 for hook in self._forward_pre_hooks.values():
324 hook(self, input)
--> 325 result = self.forward(*input, **kwargs)
326 for hook in self._forward_hooks.values():
327 hook_result = hook(self, input, result)
~/openface-pytorch/layer/spatialcrossmaplrn.py in forward(self, x)
53 # y = x / { (k + alpha / size * sum)^beta }
54 layer_square_sum = layer_square_sum * self.alpha / self.size + self.k
---> 55 output = x * torch.pow(layer_square_sum, -self.beta)
56
57 # recover 3D
RuntimeError: The size of tensor a (64) must match the size of tensor b (384) at non-singleton dimension 1
This appears to result from the batch size. If this gets changed to 1, the tensors match, but the following error appears:
TypeError Traceback (most recent call last)
<ipython-input-9-20a43768b30f> in <module>()
11
12 # CPU
---> 13 output = model(Variable(input))
14 print('CPU done')
15
~/envs/face-recognition/lib/python3.5/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
323 for hook in self._forward_pre_hooks.values():
324 hook(self, input)
--> 325 result = self.forward(*input, **kwargs)
326 for hook in self._forward_hooks.values():
327 hook_result = hook(self, input, result)
~/envs/face-recognition/lib/python3.5/site-packages/torch/nn/modules/container.py in forward(self, input)
65 def forward(self, input):
66 for module in self._modules.values():
---> 67 input = module(input)
68 return input
69
~/envs/face-recognition/lib/python3.5/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
323 for hook in self._forward_pre_hooks.values():
324 hook(self, input)
--> 325 result = self.forward(*input, **kwargs)
326 for hook in self._forward_hooks.values():
327 hook_result = hook(self, input, result)
~/openface-pytorch/layer/normalize.py in forward(self, x)
26
27 for i in range(x.size(0)):
---> 28 x[i].div_(norm.data[i][0])
29 return x
30 else:
TypeError: 'float' object is not subscriptable
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 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. 📊📈🎉
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 ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.