Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.
6404 Discussions

Toolkit Error: Internal Error of a caffe network including concatination with input

idata
Employee
771 Views

Hi,

 

The following error occurred when I tried to compile my network including a concat layer by mvNCCompile v02.00:

 

[Error 17] Toolkit Error: Internal Error: Could not build graph. Missing link: data

 

The compilation succeeded when I removed the concat layer. So I tested the following simple network and the error happened again:

 

name: "test_concat"

 

input: "data"

 

input_shape

 

{ dim:1 dim:1 dim:64 dim:64 }

 

layer {

 

name: "Convolution"

 

type: "Convolution"

 

convolution_param {

 

num_output: 16

 

kernel_size: 3

 

dilation: 1

 

pad: 1

 

}

 

bottom: "data"

 

top: "Convolution"

 

}

 

layer {

 

name: "BatchNormalization"

 

type: "BatchNorm"

 

batch_norm_param {

 

use_global_stats: true

 

}

 

param {

 

lr_mult: 0

 

}

 

param {

 

lr_mult: 0

 

}

 

param {

 

lr_mult: 0

 

}

 

bottom: "Convolution"

 

top: "BatchNormalization"

 

}

 

layer {

 

name: "ReLU"

 

type: "ReLU"

 

bottom: "BatchNormalization"

 

top: "ReLU"

 

}

 

layer {

 

name: "Concatenate"

 

type: "Concat"

 

concat_param {

 

axis: 1

 

}

 

bottom: "ReLU"

 

bottom: "data"

 

top: "Concatenate"

 

}

 

layer {

 

name: "Convolution2"

 

type: "Convolution"

 

convolution_param {

 

num_output: 1

 

kernel_size: 3

 

dilation: 1

 

pad: 1

 

}

 

bottom: "Concatenate"

 

top: "Convolution2"

 

}

 

layer {

 

name: "Score"

 

type: "Sigmoid"

 

bottom: "Convolution2"

 

top: "Score"

 

}

 

It seems the duplication of the input-data may cause the error. I could not find any restriction from documents. Is it a bug of SDK?

 

I would appreciate any help.

0 Kudos
0 Replies
Reply