I am trying to convert Tensorflow face detection uint8 model(distributed with opencv) to IR format . I am facing below error during conversion. Below is command line for model optimizer. Please help me here.
[ ERROR ]
Unexpected exception happened during extracting attributes for node flatten/Reshape/shape/_0__cf__0.
Original exception message:
Below is command line for model optimizer
python mo_tf.py --input_model C:\Users\cverma\Documents\Training\Edge_AI\project\gad\opencv_face_detector_uint8.pb --output_dir C:\Users\cverma\Documents\Training\Edge_AI\project\gad\ --reverse_input_channels --input_shape [1,300,300,3] --mean_values [104,117,123] --tensorflow_use_custom_operations_config C:\Users\cverma\Documents\Training\Edge_AI\project\gad\ssd_config\ssd_v2_support.json
I have attached tensor flow model file and below is link from where this model downloaded.
The model you are using is not of the validated TensorFlow models for OpenVINO toolkit. Please find the list of tested models here - https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_convert_model_Convert_Model_From_TensorFlow.html
However, model optimizer is able to work with unsupported models. In order to debug this further and identify unsupported operations used in your model you have to run model optimizer with the switch --log_level DEBUG
So then you'll be able to put supported operations into your model.
Please also note that we have the following validated face detection models:
IR face/age/gender recognition http://docs.openvinotoolkit.org/latest/_models_intel_index.html
Best regards, Max.