convert tf resnet50_ssd model to IR failed

[pig__dudu]

Hi,

I trained a model from the pretained model ssd_resnet_50_fpn_coco   with tensorflow=1.12.0 . the command i run is

python3 /opt/intel/openvino_2019.1.133/deployment_tools/model_optimizer/mo.py \
    --framework=tf \
    --input_model /root/sk_person_3.pb \
    --tensorflow_use_custom_operations_config /opt/intel/openvino_2019.1.133/deployment_tools/model_optimizer/extensions/front/tf/ssd_v2_support.json \
    --tensorflow_object_detection_api_pipeline_config /root/tf_frozen_models/ssd_resnet50/pipeline.config \
    --data_type FP32 \
    --reverse_input_channels \
    --input=image_tensor \
    --input_shape=[1,640,640,3] \
    --output=detection_classes,detection_scores,detection_boxes,num_detections \
    --log_level=DEBUG

but it failed.

I have attached the pb model and debug log to post. Looking forward to your help.

[pig__dudu]

additional，the error happed after i change the ssd_v2_support.json according to this post  becase i have the same error.

I try it with openvino_2019.1.144 .  it has the same error.

[HemanthKum_G_Intel]

Hi dudu,

We are working out internally to look upon this issue. Reply to you once we get an answer.

[Connolly__John]

Was there any resolution on this? I am encountering the same issue which I have posted about here https://software.intel.com/en-us/forums/computer-vision/topic/820894

[Shubha_R_Intel]

Dear Connolly, John,

I gave you a possible fix in 820894  Did you try it ? And if it still fails can you attach a full log with --log_level DEBUG to  820294?

Thanks !

Shubha

[HemanthKum_G_Intel]

Hi John,

Please, use the following command line to convert the model in Intel® Distribution of OpenVINO™ Toolkit 2019 R3:

./mo.py --input_model ./sk_person_3.pb --tensorflow_use ./ssd_support_custom.json --tensorflow_obje ./ssd_resnet_50_v1_fpn_coco.config

The customized ssd_support_custom.json file is attached.