Error while trying to convert custom trained Tensorflow SSD Mobilenet V2 model to IR format.

[dilip96]

Hi,

I am not able to convert my custom trained model to IR format. I have installed the Openvino 2020.1 version in linux 18.04 LTS system.

Following are the links that I reffered to install Openvino and conversion process.

https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_Object_Detection_API_Models.html#ssd_single_shot_multibox_detector_topologies

https://docs.openvinotoolkit.org/latest/_docs_install_guides_installing_openvino_linux.html

This is the error I am facing,

[ ERROR ]  Failed to convert tokens to dictionary: Wrong character "Set" in position 7
[ ERROR ]  Failed to generate dictionary representation of file.
[ ANALYSIS INFO ]  Your model looks like TensorFlow Object Detection API Model.
Check if all parameters are specified:
    --tensorflow_use_custom_operations_config
    --tensorflow_object_detection_api_pipeline_config
    --input_shape (optional)
    --reverse_input_channels (if you convert a model to use with the Inference Engine sample applications)
Detailed information about conversion of this model can be found at
https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_Object_Detection_API_Models.html
[ ERROR ]  Exception occurred during running replacer "ObjectDetectionAPIPreprocessorReplacement" (<class 'extensions.front.tf.ObjectDetectionAPI.ObjectDetectionAPIPreprocessorReplacement'>): Failed to parse pipeline.config file /home/labuser/Openvino_conversion/ssd_mobilenet_v2_coco.config

I have attached the complete log with the post. PFA.

Note: I am able to successfully convert Pre-Trained Zoo model, by changing the input prameters for the mo_tf.py file.

[JAIVIN_J_Intel]

Hi Dilip,

Which version of Tensorflow are you using?

Can you mention the model optimizer command used to convert the model?

Can you share the model and necessary files to reproduce the issue from our end? If required, I can send you a Private Message to share the files privately.

Regards,

Jaivin