Hi @erik

I tried to convert a tensorflow model (.pb) into a MyriadX blob using Luxonis MyriadX Blob Converter online with

Model file: middlebury_d400.pb (you may download the model using wget from here:

wget -P . -N https://storage.googleapis.com/tensorflow-graphics/models/hitnet/de
fault_models/middlebury_d400.pb

Model optimizer params:
--data_type=FP16 --input_shape "[1,480,640,6]" --disable_nhwc_to_nchw

MyriadX compile params:
-ip FP16
Shaves: 8

Then I got the following error message:
Error Message:
Command failed with exit code 1, command: /opt/intel/openvino/deployment_tools/inference_engine/lib/intel64/myriad_compile -m /tmp/blobconverter/e334657f91054914b505b1b1408c267c/middlebury_d400/FP16/middlebury_d400.xml -o /tmp/blobconverter/e334657f91054914b505b1b1408c267c/middlebury_d400/FP16/middlebury_d400.blob -c /tmp/blobconverter/e334657f91054914b505b1b1408c267c/myriad_compile_config.txt -ip FP16
Console output (stdout)

Inference Engine: 
	IE version ......... 2021.4.2
	Build ........... 2021.4.2-3974-e2a469a3450-releases/2021/4
[1;33m[Warning][VPU][Config] Deprecated option was used : VPU_MYRIAD_PLATFORM[0m

Error output (stderr)

[ GENERAL_ERROR ] 
/home/jenkins/agent/workspace/private-ci/ie/build-linux-ubuntu20/b/repos/openvino/inference-engine/src/vpu/graph_transformer/src/frontend/frontend.cpp:441 Failed to compile layer "level0/shared/prop0/WarpImageWithHypotheses/warp_and_aggregate_sad/Abs_3": unsupported layer type "Abs"

Then I check VPU supported layers here:
And found "Abs" is supported by VPU

Could you please help to see what am I missing? Thanks a lot.

  • erik replied to this.

    Hello ynjiun ,
    Even though it says it's supported on OpenVINO docs, I don't think it actually is. Eg. see the discussion here. It might be best if you open an issue on OpenVINO github regarding this problem. But as mentioned in the link, I would switch the layers as PINTO suggested: Abs <-> Sqrt(Multiply(X * X))
    Thanks, Erik