Note: This project is not maintained anymore. Mobilenet implementation is already included in Keras Applications folder. Mobilenet

Keras MobileNet

Google MobileNet Implementation using Keras Framework 2.0

Project Summary

• This project is just the implementation of paper from scratch. I don't have the pretrained weights or GPU's to train :)

• Separable Convolution is already implemented in both Keras and TF but, there is no BN support after Depthwise layers (Still investigating).

• Custom Depthwise Layer is just implemented by changing the source code of Separable Convolution from Keras. Keras: Separable Convolution

• There is probably a typo in Table 1 at the last "Conv dw" layer stride should be 1 according to input sizes.

• Couldn't find any information about the usage of biases at layers (not used as default).

TODO

•  Add Custom Depthwise Convolution

•  Add BN + RELU layers

•  Check layer shapes

•  Test Custom Depthwise Convolution

•  Benchmark training and feedforward pass with both CPU and GPU

•  Compare with SqueezeNet

Library Versions

• Keras v2.0+

• Tensorflow 1.0+ (not supporting Theano for now)

References

1. Keras Framework

2. Google MobileNet Paper

Licence

MIT License

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