Caffe -

[Julien Guillaumin]

Hi,

I'm looking to get back trained weights from a specific layer in a network (or a subset of layers, like an inception module) to re-use it after as initial weights in a layer from a new network . The goal is to construct a hierarchical classifier, where the training follows the hierarchy.
But I'm new with Caffe ...
Weights are stored in .caffemodel for inference, but can we extract some values from this file ?

With Theano, it's possible to get back weights from a specific layer after training, as a numpy array after a training phase.

Thank for your help !

Julien

[Evan Shelhamer]

Hey,

There are two examples from the homepage that show just that:

- inspecting net outputs and parameters http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/00-classification.ipynb

- loading and editing model parameters http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/net_surgery.ipynb

Happy brewing,

Evan Shelhamer

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users+unsubscribe@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/1f624ee6-3a7a-4334-9efb-cba145dee926%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Julien Guillaumin]

Thank you very much ! 

It works well. 

Le samedi 8 octobre 2016 03:07:21 UTC+2, Evan Shelhamer a écrit :

Hey,

There are two examples from the homepage that show just that:

- inspecting net outputs and parameters http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/00-classification.ipynb

- loading and editing model parameters http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/net_surgery.ipynb

Happy brewing,

Evan Shelhamer

On Fri, Oct 7, 2016 at 7:46 AM, Julien Guillaumin <julien.gui...@gmail.com> wrote:

Hi,

I'm looking to get back trained weights from a specific layer in a network (or a subset of layers, like an inception module) to re-use it after as initial weights in a layer from a new network . The goal is to construct a hierarchical classifier, where the training follows the hierarchy.
But I'm new with Caffe ...
Weights are stored in .caffemodel for inference, but can we extract some values from this file ?

With Theano, it's possible to get back weights from a specific layer after training, as a numpy array after a training phase.

Thank for your help !

Julien

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users...@googlegroups.com.