Cannot convert faster rcnn inception resnet v2 tensorflow 2 model to IR.

[simranS]

Hi,

I tried converting a tensorflow 2 model Faster R-CNN Inception ResNet V2 1024x1024 from model zoo using command mentioned in 

https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_TensorFlow.html#Convert_From_TF2X :

Command: python3 mo_tf.py --saved_model_dir /home/ubuntu/Desktop/tensorflow_2_model/content/fine_tuned_model/saved_model/ --tensorflow_object_detection_api_pipeline_config /home/ubuntu/Desktop/tensorflow_2_model/content/fine_tuned_model/pipeline.config --tensorflow_use_custom_operations_config /opt/intel/openvino_2020.4.287/deployment_tools/model_optimizer/extensions/front/tf/faster_rcnn_support.json --reverse_input_channels --input_shape [1,800,1333,3] 

It gives the following error:

[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID"

------------------------------------------------------------------------------------------------------------

I was able to convert the mask_rcnn_inception_v2_coco_2018_01_28, which I downloaded from open model zoo.

Are faster rcnn tensorflow 2 models supported in Model Optimizer for IR conversion?

[AthiraM_Intel]

Hi,

Thanks for reaching out to us.

Since your issue is related to OpenVINO, we are moving this query to OpenVINO forum for a faster response.

[simranS]

Hi, 

Can you provide the link where this issue is oded. So that I can track it.

Thanks

[Iffa_Intel · Accepted solution]

Greetings,

TF2 does supported in MO for IR convertion. You can refer here for further details: https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_TensorFlow.html#Convert_From_TF2X

Officially supported Topologies are also mentioned in above documentation.

I suggest you try with the Officially supported topologies: faster_rcnn_inception_resnet_v2_atrous_coco_2018_01_28.tar.gz OR faster_rcnn_inception_resnet_v2_atrous_lowproposals_coco_2018_01_28.tar.gz

Just to ensure that your toolkit is correctly setup and usable.

Sincerely,

Iffa

[simranS]

Yes I was able to convert the above mentioned models to IR . Does this mean I should be using the atrous version of faster rcnn to retrain model outside openvino as the IR conversion for these is supported?

[Iffa_Intel]

You can use the readily available or you can train your own model, as long as it is supported by openvino as stated by the documentation that I gave to you previously.

Sincerely,

Iffa

[Iffa_Intel]

Greetings,

Intel will no longer monitor this thread since this issue has been resolved. If you need any additional information from Intel, please submit a new question. 

Sincerely,

Iffa