Conversion of ssd_resnet50_v1_fpn_coco model using 2019 R1.1

[morita_1]

I tried to convert the custom ssd_resnet50_v1_fpn_coco into IR format files in order to use it in LabVIEW. But the error happened.

[ ERROR ] Cannot infer shapes or values for node "Postprocessor/Cast_1".

Models should be compatible with OpenVINO 2019 R1.1.

How can I convert the model ?

Thank you.

TensorFlow version 1.14.0

Model Optimizer version: 2019.1.1-83-g28dfbfd

[Iffa_Intel]

Greetings,

ResNet is a Tensorflow's topology which is supported by OpenVINO. It is listed here.

You may also refer to that documentation for the conversion steps.

Those that are listed there are validated to be working.

Sincerely,

Iffa

[morita_1]

Thank you for your reply.

I can convert the TensorFlow Object Detection Model Zoo ssd_resnet50_v1_fpn_coco with referring to the conversion steps.
However I cannot convert custom model trained with TensorFlow Object Detection API into IR format.

I use the following conversion command.

python mo_tf.py --input_model D:/ModelDirectory/frozen_inference_graph.pb --tensorflow_use_custom_operations_config extensions/front/tf/ssd_v2_support.json --tensorflow_object_detection_api_pipeline_config D:/ModelDirectory/pipeline.config --output_dir D:/Model Directory

[Iffa_Intel]

Make sure that your topology used with the Tensorflow model is listed here.

If it is not listed, your topology is not supported by OpenVINO, and failure in conversion is expected.

Sincerely,

Iffa

[Iffa_Intel]

Greetings,

Intel will no longer monitor this thread since we have provided a solution. If you need any additional information from Intel, please submit a new question.

Sincerely,

Iffa