rgmyr / keras-texture Goto Github PK

File "C:\Users\chouchou\Documents\Python Scripts\BilinearCNN\networks\dep_net.py", line 63, in dep_net
output_head = BilinearModel()([encode_head]) # flat, l2_normalized output by default
File "C:\Users\chouchou\Anaconda3\lib\site-packages\keras\engine\base_layer.py", line 431, in call
self.build(unpack_singleton(input_shapes))
File "C:\Users\chouchou\Documents\Python Scripts\BilinearCNN\layers\bilinearmodel.py", line 31, in build
self._shapecheck(input_shape)
TypeError: _shapecheck() takes 1 positional argument but 2 were given

I am sure some mistakes are committed.

Since the last push I've worked on this codebase inside of a different private project. I've corrected a couple of mistakes and fixed a few bugs there, which I should pull back into this package.

Just making a public note here as motivation to actually get around to doing that.

Another question is as follows, when I was trying deepten network. I met a question when creating an residual encoding layer with learnable codebooks.
The error is as follows:
/layers/encoding.py,line 69 in build
self.D = input_shape.as_list()[-1]
AttributeError:'tuple' object has no attribute 'as_list'
swig/python detected a memory leak of type 'int64_t *'

Hello,
I'm trying to use the encoding layer from DeepTEN example.Even with the simple network architecture my model(network+encoding layer) starts overfitting. When I run my model without encoding it doesn't overfit then.Any help is appreciated.
Thanks!!

Great job！
But I have a question about the codebook .In my understanding, dictionary learning is equal to sparse encoding, but why is the size of the codebook is （32,128）？thank you very much！

When I was trying to use the deepten network, and I find out that there is no model named TextureModel in the model module. Maybe it's because the name for the model is wrong.