simple tool to translate caffe model to keras

It is inspired by:

• keras's caffe branch
• https://github.com/pfnet/chainer/blob/master/chainer/functions/caffe (functions and examples)
• https://gist.github.com/baraldilorenzo (VGG16 model for Keras)
• https://github.com/mitmul/chainer-imagenet-vgg

Example:

please browse the examples directory

Dependencies:

• Python:

  • keras
  • google.protobuf

• caffe:

  • caffe's python lib (caffe/python)

Some Problems:

• 1.Some function in load_caffe_params and save_weights which I think is Inefficient.

• 2.The Proto2Keras is combine keras's caffe branch and chainer's caffe function, which some of them unfamiliar for me. So I just keep it.

• 3.The last and most import is I test this in a very very few example, and certainly there are a lot of bugs. So feel free to contact me(heyflypigATgmail.com), I am a newbie to DeepLearning.