Issue with converting faster RCNN ONNX model to IR

[Pruthvin]

Dear Team,

We are facing issues while converting the faster rcnn ONNX model to IR. PFB commands used  and errors faced.

COMMAND

python3 /opt/intel/openvino_2021.3.394/deployment_tools/model_optimizer/mo.py --input_model /home/pruthvin/SteelDefect_FRCNN_ModelFiles/SteelDefect_FRCNN_ONNXModel/SteelDefect_FRCNN_ONNXModel.onnx --input_shape [1,1,1024,512] --transformations_config /opt/intel/openvino_2021.3.394/deployment_tools/model_optimizer/extensions/front/onnx/faster_rcnn.json --output_dir /home/pruthvin/SteelDefect_FRCNN_ModelFiles/SteelDefect_FRCNN_PBFiles --model_name tf_v3 --reverse_input_channels

ERROR

[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.load.onnx.loader.ONNXLoader'>): Unexpected exception happened during extracting attributes for node Size__1200.
Original exception message: Size has mandatory `output_type` attribute

OPENVINO VERSION

openvino_2021.3.394

UBUNTU VERSION

Ubuntu 18.04.4 LTS

CPU

Intel(R) Xeon(R) Platinum 8180 CPU @ 2.50GHz

Regards,

Pruthvin S

[IntelSupport]

Hi Pruthvin,

Thanks for reaching out. Which specific OpenVINO models that you are using? Could you please share your models and any related files so that we can test them on our side?

Regards,

Aznie

[Pruthvin]

Dear Team,

We cannot share the model as it is specific to customer, kindly provide the links or solutions related to the posted error which would help us in landing at the solution.

Regards,

Pruthvin S

[IntelSupport]

Hi Pruthvin S,

Could you try to use the parameters from the Convert ONNX* Faster R-CNN Model to the Intermediate Representation documentation as mention in Step 2? Please give it a try and get back to me with the result. Your error might due to the unsupported node/layer or incompatibility input size since you are using the custom model.

Regards,

Aznie

[Syamimi_Intel]

Hi Pruthvin S,

Thank you for your question. If you need any additional information from Intel, please submit a new question as this thread is no longer being monitored.

Regards,

Syamimi