Two top blobs in train.prototxt when using caffe.NetSpec() in Python

Question

I am using caffe.NetSpec in python to define and export in architecture using the following code :

conv1_1 = L.Convolution(data,top='conv1_1',name='conv1_1',
                        convolution_param=
                        {'kernel_size':3,'num_output':64,'pad':1},
                        param=[{'lr_mult':1, 'decay_mult':1},
                               {'lr_mult':2,'decay_mult':0}])

But, when generating the train.protxt, two Top blobs appears in the layer as follow :

layer {
name: "conv1_1"
type: "Convolution"
bottom: "Data1"
top: "Convolution1"
top: "conv1_1"
param {
  lr_mult: 1
  decay_mult: 1
}
param {
  lr_mult: 2
  decay_mult: 0
}
convolution_param {
  num_output: 64
  pad: 1
  kernel_size: 3
  }
}

What's going wrong here? Thanks

@Shai: I'm just stacking a bunch of layers and then feed the last layer to the 'to_proto' function. It's a very deep network. You need all the code? Note that the same behavior happens for all layers. — Saeed
– Saeed, Commented Feb 28, 2017 at 6:23

Shai · Accepted Answer · 2017-02-28 06:27:56Z

1

How about using a NetSpec() object?

import caffe

ns = caffe.NetSpec() # use this object to store the layers
ns.data, ns.label = L.Data(name='data',  ntop=2, 
                           data_param={'source':'', 'batch_size': 32})
ns.conv1_1 = L.Convolution(ns.data, name='conv1_1',
                    convolution_param=
                    {'kernel_size':3,'num_output':64,'pad':1},
                    param=[{'lr_mult':1, 'decay_mult':1},
                           {'lr_mult':2,'decay_mult':0}])
print str(ns.to_proto()) # print the net stored in ns object

answered Feb 28, 2017 at 6:27

Shai

115k39 gold badges259 silver badges398 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Scott Yang Over a year ago

In this way, can we use variables inside the layer 'top's? For example, ns.conv1_1, ns.conv1_2 etc. We don't wanna write it so many times, right?

Collectives™ on Stack Overflow

Two top blobs in train.prototxt when using caffe.NetSpec() in Python

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related