[Model Optimizer Error] Freeze Mask RCNN tensorflow model

[Jeff__Liao]

• OS: Windows 10 pro
• CPU: Intel(R) Core(TM) i5-7500 CPU @3.40GHz
• GPU: None
• OpenVINO: 2020.1
• python: 3.7.6
• tensorflow: 1.15.0

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

Hi everyone, 

When wanting to use "mask_rcnn_resnet50_atrous_coco_2018_01_28" from model zoo and freezing it, I met a problem below:

There are my input arguments

• --input_model :  <folder path>/frozen_inference_graph.pb
• --output_dir : <folder path>
• --input_checkpoint : <folder path>/checkpoint
• --tensorflow_use_custom_operations_config : <folder>/mask_rcnn_support.json
• --tensorflow_object_detection_api_pipeline_config : <folder>/pipeline.config

Error :

Anyone who also meets the same problem can help resolve it. Thanks !!

I have debugged on the model convert several days ....

[JAIVIN_J_Intel]

Hi Liao,

I have just tried to convert this model using model optimizer and was able to successfully obtain the IR using the follwing command:

python3 /opt/intel/openvino/deployment_tools/model_optimizer/mo_tf.py  --input_model <path-to-model>/frozen_inference_graph.pb --tensorflow_use_custom_operations_config /opt/intel/openvino/deployment_tools/model_optimizer/extensions/front/tf/mask_rcnn_support.json --tensorflow_object_detection_api_pipeline_config  <path-to-model>/pipeline.config -o <path-to-save-output>

Please try again after removing the --input_checkpoint argument.

Regards,

Jaivin

[Jeff__Liao]

Hi Jaivin,

I have tried what you said. But still cannot work and output the same error.

There is my input argument below:

python mo_tf.py --input_model C:/Users/<username>/Downloads/Release/mask_rcnn_inception_v2_coco_2018_01_28-1/inference_graph/frozen_inference_graph.pb --output_dir C:/Users/<username>/Downloads/Release/mask_rcnn_inception_v2_coco_2018_01_28-1/inference_graph/ --tensorflow_use_custom_operations_config C:/"Program Files (x86)"/IntelSWTools/openvino_2020.1.033/deployment_tools/model_optimizer/extensions/front/tf/mask_rcnn_support.json

I run it with --log_level DEBUG and show these message.

By the way, I run mo_tf.py under the virtual environment. I don't know that it's a problem i meet

Best Regards.

Jeff

[JAIVIN_J_Intel]

Hi Jeff,

On the command you have mentioned, --tensorflow_object_detection_api_pipeline_config parameter has not been used. Please use the parameter and try again.

Regards,

Jaivin