Community
cancel
Showing results for 
Search instead for 
Did you mean: 
58 Views

Kernel-x error when moving from OpenVino R3 to R4

hi,

I tried converting a network using the following instructions:

https://software.intel.com/en-us/articles/OpenVINO-ModelOptimizer

I compile it using the following command:

${INTEL_CVSDK_DIR}/deployment_tools/model_optimizer/mo.py  --input_model det1test.caffemodel --output_dir FP16test --data_type FP16

it works fine with R3 but when I compile it with R4 I get the following error:

 self.pnet.reshape({self.input_blob_pnet: (tmpshape[0], tmpshape[1], ws, hs)})
 File "ie_api.pyx", line 281, in inference_engine.ie_api.IENetwork.reshape
RuntimeError: Failed to infer shapes for Convolution layer with error: No such parameter name 'kernel-x' for layer 
/teamcity/work/scoring_engine_build/releases_openvino-2018-r4/include/details/ie_exception_conversion.hpp:81

It looks like the inference engine is looking for layers that don't exist. when I diff the two networks I can see that those layers are missing or renamed, here are some diffs:

R3

<?xml version="1.0" ?>
<net batch="1" name="PNet" version="2">

R4

<?xml version="1.0" ?>
<net batch="1" name="PNet" version="3">

R3

<data dilations="1,1" group="1" kernel="3,3" output="10" pads_begin="0,0" pads_end="0,0" strides="1,1"/>

R4

<data dilation-x="1" dilation-y="1" group="1" kernel-x="3" kernel-y="3" output="10" pad-b="0" pad-r="0" pad-x="0" pad-y="0" stride="1,1,1,1" stride-x="1" stride-y="1"/>

Any ideas what is going on and how to fix this?

 

 


 

0 Kudos
0 Replies
Reply