Multi-scale network
25 views
Skip to first unread message
Anton Varfolomeev
Jul 21, 2014, 8:06:02 AM7/21/14
to ebl...@googlegroups.com
Does anybody know how to implement in EBLearn multi-scale network (that uses outputs of different layers) ?
Like one that was used in (Sermanet, Kavukcuoglu, and LeCun 2011), (Sermanet and LeCun 2011).
It seems that one must combine independent 'convolution-sub-sampling' and 'pass-through' modules in on layer - but
I can't catch the proper keywords
aravi...@gmail.com
Sep 21, 2017, 4:05:03 PM9/21/17
to eblearn
On Monday, July 21, 2014 at 5:36:02 PM UTC+5:30, Anton Varfolomeev wrote:
> Hi, all.
>
>
> Does anybody know how to implement in EBLearn multi-scale network (that uses outputs of different layers) ?
> Like one that was used in (Sermanet, Kavukcuoglu, and LeCun 2011), (Sermanet and LeCun 2011).
>
>
>
>
>
> Hi, all.
>
>
> Does anybody know how to implement in EBLearn multi-scale network (that uses outputs of different layers) ?
> Like one that was used in (Sermanet, Kavukcuoglu, and LeCun 2011), (Sermanet and LeCun 2011).
>
>
>
>
>
> It seems that one must combine independent 'convolution-sub-sampling' and 'pass-through' modules in on layer - but
> I can't catch the proper keywords
I'm not familiar with EBLearn configuration files. But may be svhn.conf would help?
> I can't catch the proper keywords
http://svn.code.sf.net/p/eblearn/code/trunk/demos/svhn/svhn.conf
On the other hand is it possible to obtain the configuration files used for GTSRB? Specially the 'EBLearn 2LConvNet ms 108 feats' model
Reply all
Reply to author
Forward
0 new messages