https://github.com/titu1994/keras-octconv
Keras implementation of Octave Convolutions
https://github.com/titu1994/keras-octconv
Last synced: 3 months ago
JSON representation
Keras implementation of Octave Convolutions
- Host: GitHub
- URL: https://github.com/titu1994/keras-octconv
- Owner: titu1994
- License: mit
- Created: 2019-04-20T03:35:25.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-04-23T01:31:47.000Z (over 6 years ago)
- Last Synced: 2025-03-18T09:02:34.762Z (7 months ago)
- Language: Python
- Size: 127 KB
- Stars: 53
- Watchers: 4
- Forks: 17
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Keras Octave Convolutions (OctConv)
-----
Keras implementation of the Octave Convolution blocks from the paper [Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution](https://arxiv.org/abs/1904.05049).
# Usage
`Octave Convolutions` are a semi-drop-in-replacement for regular convolution layers.
They are implemented in 3 major steps:
## Intiailization of Dual Path Flow
Use the `initial_octconv` block from `octave_conv.py` to initialize the Octave convolution blocks. This function accepts a single input tensor, and returns **two** output tensors : The high frequency pathway and low frequency pathway tensors, in that order
```python
ip = Input(...)
x_high, x_low = initial_conv(ip, ...)
```
## Add any number of Octave Convolution Blocks
Once the two frequency pathways have been obtained, use any number of `octconv_block` from `octave_conv.py` to make the network larger.
**NOTE**:
Each of these blocks accept *two* input tensors, and emits *two* output tensors.
```python
x_high, x_low = octconv_block(x_high, x_low, ...)
x_high, x_low = octconv_block(x_high, x_low, ...)
x_high, x_low = octconv_block(x_high, x_low, ...)
```
## Merging the streams back together
Once you are finished adding `octconv_block`s, merge the two frequency pathways using `final_octconv` from `octave_conv.py`.
This block accepts *two* input tensors and *one* output tensor.
```python
x = final_octconv(x_high, x_low, ...)
...
```
## Acknowledgements
This code is heavily based on the MXNet implementation by [terrychenism](https://github.com/terrychenism) at https://github.com/terrychenism/OctaveConv.
# Requirements
----
- Keras 2.2.4+
- Tensorflow 1.13+ (2.0 support depends on when Keras will support it) / Theano (not tested) / CNTK (not tested)