No translator found for TFLite_Detection_PostProcess node

[RySAP]

HI,

I tried to execute exported Efficientdet.

Converting to IR file is done by referring following tutorial.

https://docs.openvino.ai/2024/notebooks/002-openvino-api-with-output.html#tensorflow-lite-model

I got below error message in my program.

(openvino2) ubuntu@ubuntu-DH610:~/openvino$ python convert.py
Traceback (most recent call last):
File "/home/ubuntu/openvino/convert.py", line 8, in <module>
model_tflite = core.read_model(tflite_model_path)
File "/opt/intel/openvino_2023.3.0/python/openvino/runtime/ie_api.py", line 511, in read_model
return Model(super().read_model(model))
RuntimeError: Exception from src/inference/src/core.cpp:85:
Check　'translate_map.count(decoder->get_op_type())'failed at src/frontends/tensorflow_lite/src/frontend.cpp:233:
FrontEnd API failed with OpConversionFailure:
No translator found for TFLite_Detection_PostProcess node.

Thanks,

[Iffa_Intel]

Hi,

from my side, I managed to convert a model using the method in the link you mentioned :

Could you share:

1. Your Ubuntu version

2. Steps that you did till point of errors

3. Could you share that Efficientdet model?

4. Any custom changes that you did to the model?

I recommend you install OpenVINO PyPI and use the omz_downloader and omz_converter instead.

Especially if the model you mentioned is an OpenVINO pre-trained model.

Download the model:

Convert the model:

Cordially,

Iffa

[RySAP]

Hi,

Thank you for reply.

1. Your Ubuntu version  22.04

   OpenVINO Version    2023.3 LTS

2. Steps that you did till point of errors

1783 mv l_openvino_toolkit_ubuntu22_2023.3.0.13775.ceeafaf64f3_x86_64 /opt/intel/openvino_2023.3.0
1784 sudo mv l_openvino_toolkit_ubuntu22_2023.3.0.13775.ceeafaf64f3_x86_64 /opt/intel/openvino_2023.3.0
1785 cd /opt/intel/openvino_2023.3.0/
1786 sudo -E ./install_dependencies/install_openvino_dependencies.sh
1787 less ./python/requirements.txt
1788 python 3 -m pip install -r ./python/requirements.txt
1789 python3 -m pip install -r ./python/requirements.txt
1790 cd /opt/intel/
1791 sudo ln -s openvino_2023.3.0/ openvino_2023
1792 source /opt/intel/openvino_2024/setupvars.sh
1793 cd openvino_2023
1794 source /opt/intel/openvino_2024/setupvars.sh
1795 source /opt/intel/openvino_2023/setupvars.sh
1796 sudo vi ~/.bashrc
1797 python -m pip install --upgrade pip
1798 python3 -m pip install --upgrade pip
1799 ls
1800 mkdir pyenv
1801 cd pyenv/
1802 python3 -m venv --system-site-packages openvino
1803 source openvino/bin/activate
1804 pip list
1805 python -m pip install opencv-python
1806 cd ~
1807 mkdir openvino
1808 cd openvino/
1809 vi deivce_confirm.py
1810 mv deivce_confirm.py device_confirm.py
1811 python device_confirm.py
1812 mkdir models
1813 cd models/
1814 cp ~/models/efficientdet_lite3_bs6_f16_20230718.tflite
1815 cp ~/models/efficientdet_lite3_bs6_f16_20230718.tflite .
1816 ls
1817 cd ..
1818 vi convert.py
1819 python convert.py
1820 vi convert.py
1821 python convert.py
1822 vi convert.py
1823 ls
1824 vi convert.py
1825 python convert.py
1826 vi download_tflite.py
1827 python download_tflite.py
1828 vi download_tflite.py
1829 python download_tflite.py
1830 vi download_tflite.py
1831 python download_tflite.py
1832 python -m pip install requests tqdm ipywidgets
1833 python download_tflite.py
1834 ls models/
1835 ls
1836 vi convert.py
1837 python convert.py

3. Could you share that Efficientdet model?

The Efficientdet model is from TensorFlow Lite Model Maker library.

It is converted from TensorFlow to TensorFlow Lite by Float16.

4. Any custom changes that you did to the model?

I Attached.

Thanks,

[Iffa_Intel]

Hi,

if the model is not an Intel pre-trained model, could you share the model also, for us to test?

Cordially,

Iffa

[RySAP]

Hi,

I tried two models I attached.

Thanks

[RySAP]

Sorry. I attached.

[Iffa_Intel]

Thanks, we are investigating this.

Cordially,

Iffa

[RySAP]

How is the state of progress after that?

[Iffa_Intel]

Hi,

We are still investigating this and we are getting some final clarification for this issue.

We appreciate your patience.

Thanks.

Cordially,

Iffa

[RySAP]

Hi,

Thank you for your effort.

I got answer from another intel support. 

They found it is bug of conversion with efficientdet TFLite models.  They sent their analysis to engineering team.

Thanks,

[Iffa_Intel]

HI,

glad to know that.

Intel will no longer monitor this thread since this issue has been resolved. If you need any additional information from Intel, please submit a new question. 

Cordially,

Iffa