Model Optimizer ONNX conversion error

[yusuke-konno]

Hello Intel Developer,

I train deeplab v3+ with Resnet-18 on MATLAB (deeplabv3plusLayers()) and export to ONNX format.

While conversion, Model Optimizer outputs following error.

Which is wrong, MATLAB's ONNX or Model Optimizer ?

I attach compressed ONNX file.

Thank you.

----

python "C:\Program Files (x86)\IntelSWTools\openvino\deployment_tools\model_optimizer\mo.py" --steps --input_model deeplab.onnx
Model Optimizer arguments:
Common parameters:
        - Path to the Input Model:      deeplab.onnx
        - Path for generated IR:        .
        - IR output name:       deeplab
        - Log level:    ERROR
        - Batch:        Not specified, inherited from the model
        - Input layers:         Not specified, inherited from the model
        - Output layers:        Not specified, inherited from the model
        - Input shapes:         Not specified, inherited from the model
        - Mean values:  Not specified
        - Scale values:         Not specified
        - Scale factor:         Not specified
        - Precision of IR:      FP32
        - Enable fusing:        True
        - Enable grouped convolutions fusing:   True
        - Move mean values to preprocess section:       False
        - Reverse input channels:       False
ONNX specific parameters:
Model Optimizer version:        2019.3.0-408-gac8584cb7
[ INFO ] Model loading step
[ INFO ] Front phase execution step
[ INFO ] Middle phase execution step
[ ERROR ]  After partial shape inference were found shape collision for node Add2_ (old shape: [  1  64 128 128], new shape: [  1  -1 128 128])

[yusuke-konno]

I solved conversion error.

My onnx does not contain batch normalization layer's parameters.

Because MATLAB does not support intermediate result for BN layer.

See MATLAB Answers

Thank you

[HemanthKum_G_Intel]

Hi yusuke-konno,

Thank you for reporting back to OpenVINO forum.