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

Unsupported primitive of type: DetectionOutput

hsing__hj
Beginner
494 Views

Hi,

I am runing this tutorials https://github.com/intel-iot-devkit/computer-vision-inference-tutorials/tree/master/1-object-detection-ssd with Computer Vision SDK 2018.0.234.

When I execute ./IEobjectdetection I got an error like following:

        API version ............ 1.0
        Build .................. lnx_20180314
        Description ....... MKLDNNPlugin
Network loaded.
Batch size = 1
Setting-up input, output blobs...
Loading model to plugin...
Error loading model into plugin: Unsupported primitive of type: DetectionOutput name: detection_out
/teamcity/work/scoring_engine_build/releases/cvsdk-2018-r1/src/mkldnn_plugin/mkldnn_node.cpp:212

How to solve it? Thanks.

Best regards,

Hj

0 Kudos
1 Reply
Seunghyuk_P_Intel
494 Views

Hi Hj,

First of all, please take a look at "$INTEL_CVSDK_DIR/deployment_tools/inference_engine/samples/object_detection_sample_ssd" instead of web link.

* How to build

  1. copy "samples" folder to user directory, ex: ~/temp/samples
  2. ~/temp/samples$ source $INTEL_CVSDK_DIR/bin/setupvars.sh
  3. ~/temp/samples$ mkdir build; cd build; cmake .. -DCMAKE_BUILD_TYPE=Release; make
  • Then, you will see output in ~/temp/samples/build/intel64/Release/

* How to run

  1. Here is one thing you should know. mklDNN plugin (for CPU) does not support extended layer and support only standard layers. You should create extensions of custom layers as plugin and give the path/file name information to inference engine.
  2. Please check "$INTEL_CVSDK_DIR/deployment_tools/inference_engine/samples/extension/" sample. It shows how to create extension plugin and you can try with this sample. In case of "detection_out", it is implemented in this "extension" sample.
  3. If you finish build samples, "libcpu_extension.so" will be created in ~/temp/samples/build/inte64/Release/lib
  4. you should pass this path/filename to object_detection_sample_ssd from command line. 
$ source /opt/intel/computer_vision_sdk/bin/setupvars.sh
$ ./opject_detection_sample_ssd -l ~/temp/samples/build/intel64/Release/lib/libcpu_extension.so -m < model> -i <input media> 

And please take a look at how to set extension plugin in source code as well.

Regards,

Peter.

0 Kudos
Reply