[Facenet] Model optimizer convert tensorflow model failed

[Wu__David]

Hi Sir,

We are trying to convert facenet tensorflow model, but we got an error:

[ 2018-06-20 14:44:04,446 ] [ ERROR ] [ infer:112 ]  Cannot infer shapes or values for node "batch_join/fifo_queue".

[ 2018-06-20 14:44:04,446 ] [ ERROR ] [ infer:113 ]  ndarray was 1 bytes but TF_Tensor was 126 bytes

I checked others posts and found a relative topic said:

"According to the release notes, Model Optimizer for Tensorflow supports very limited list of topologies: VGG-16/19, Inception v1/v3/v4, ResNet v1 50/101/152. Facenet you shared has a lot of unsupported primitives like fifo_queue, phase_train placeholder, etc."

https://software.intel.com/en-us/forums/computer-vision/topic/742898

Is this the reason why it converted the model failed? Or it already fixed in new release.

The command we used:  python3 ...mo_tf.py --input_model 20180402-114759.pb  --output_dir ~/model/ --log_level DEBUG

The model we used is from: https://github.com/davidsandberg/facenet

Thanks.

David

[Jeffrey_M_Intel1]

Sorry for the delayed reply.  Updates to OpenVINO have been made to support Tensorflow Facenet.   Support for Facenet will be available in the next release.

[Wu__David]

Hi Jeffrey,

Glad to hear the news! Thanks for the help.

David