mvoelk / keras_layers Goto Github PK

Various Additional Keras Layers: SparseConv2D, PartialConv2D, GroupConv2D, DeformableConv2D, DepthwiseConv2D, MaxPoolingWithArgmax2D, MaxUnpooling2D, AddCoords2D, Blur2D, LayerNormalization, InstanceNormalization, WeightNormalization, EqualizedLearningRates

License: MIT License

Python 3.15% Jupyter Notebook 96.85%

keras_layers's Issues

NotImplementedError: Layer AddCoords2D has arguments in `init` and therefore must override `get_config`.

Hi, thank you for your implementation. I added AddCoords2D to my model, however, when weights are being saved, it gives an error:
NotImplementedError: Layer AddCoords2D has arguments in initand therefore must overrideget_config.

SparseConv2D Speed?

Hey, thanks for the implementation - works well, also the example. My Question is, is there a speed gain vs the common Conv2d?

DeformableConv2D not memory efficient?

Hi, thank you for your implementation. I tried to change 8 Conv2D layers with DeformableConv2D and during memory allocating (model build), GPU soon runs out of memory (8GB). Although with a small number of filters everything seems to be good. Is the DeformableConv2D operation not very memory efficient?

Recommend Projects