Skip to content
ShuffleNet Implementation using Keras Functional Framework 2.0
Python
Branch: master
Clone or download
Latest commit ae2600b Dec 8, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
weights added weights for groups 3 but not thats not the accuracy of the paper Dec 8, 2017
.gitignore first commit for shufflenet implementation in keras 2 Oct 20, 2017
.travis.yml
README.md updated weight link Dec 8, 2017
model.png added model plot Nov 11, 2017
shufflenet.py bug in number of channels per shuffleunit Nov 11, 2017
test.py removed model_plot from test.py Oct 20, 2017
train_shufflenet.py implementation learn rate deacy every 30 epochs, changed optimizer Nov 11, 2017

README.md

Keras ShuffleNet Build Status

ShuffleNet Implementation using Keras Functional Framework 2.0

There is a weight file for groups=3 but I was not able to reproduce the accuracy from the paper. Current accuracy for ImageNet is 0.5228.

Library Versions

  • Keras v2.0.8+
  • Tensorflow 1.0+

References

  1. Keras Framework

  2. ShuffleNet Paper

Licence

MIT License

Note: If you find this project useful, please include reference link in your work.

You can’t perform that action at this time.