Now I have trained GoogleNet and VGG16 respectively for the same task, i.e. googlenet.caffemodel and vgg16.caffemodel. Now I want to merge the two networks to form an ensemble model. That is, I want to define the two network in a single prototxt. This means I need to finetune on two models. How can I do that?
Asked
Active
Viewed 624 times
0
-
Possible duplicate of [Multiple pretrained networks in Caffe](http://stackoverflow.com/questions/34466901/multiple-pretrained-networks-in-caffe) – Shai Mar 02 '16 at 14:58
-
1Define the GoogLeNet and VGG16 in a single prototxt, and load their learned parameters via [net surgery](https://github.com/BVLC/caffe/blob/master/examples/net_surgery.ipynb). – Jon Mar 04 '16 at 02:54