Custom YoloV3 Model Loading Fails

[ybai]

I was blocked when trying to load a custom data trained yolov3 model based on (https://github.com/YunYang1994/tensorflow-yolov3.git) 

 

Modle files are in the attached files.

BTW the recommended implementation of yolov3 mentioned in this ticket cannot solve my problem, we need a flexible model implementation approach for high expandability.

Thanks,

Yu

[ybai]

 

[ybai]

Modle files are in the attached files.

BTW the recommended implementation of yolov3 mentioned in this ticket cannot solve my problem, we need a flexible model implementation approach for high expandability.

Thanks,

Yu

[IntelSupport]

Hi Yu,

Thanks for reaching out.

Could you share more regarding your system environment? What is your OpenVINO version and are you running on NCS or CPU?

Meanwhile, please share your model optimization command and the .pb file that you used.

Regards,

Aznie

[ybai]

Hi Aznie,

We run OpenVINO2019.3 on NCS. Since the onnx file (200M+) exceeds the attach file limitation, you can download it here

```

python /opt/intel/openvino/deployment_tools/model_optimizer/mo_onnx.py \
--input_model /home/admin/yolov3_chefcapDetect_v1.onnx \
--input_shape [1,608,608,3] \
--data_type FP16

```

Thanks,

Yu

[IntelSupport]

Hi Yu,

Thanks for the model. I have tested your model on Benchmark application and everything seems fine. Here is how I convert the model into the Intermediate Representation (IF) format:

Pyhton mo.py --input_model "Downloads\yolov3_chefchapDetect_v1.onnx" ---input_shape [1,608,608,3]

--data_type FP16 --reverse_input_channel

Then, I tested on Benchmark_app.py use this command:

benchmark_app.py -m "<INSTALL_DIR>\openvino_2021.3.394\deployment_tools\model_optimizer\yolov3_chefcapDetect_v1.xml" -d MYRIAD

Meanwhile, I would recommend you upgrade your OpenVINO to our latest version (2021.3) for better features supportability.

Regards,

Aznie

[ybai]

Hi Aznie,

Thanks for your investigation, the benchmark works fine for me, however, when I run the model on production env we caught this fault.

Regards,

Yu

[IntelSupport]

Hi Yu,

Are you using the 2019.3 version of OpenVINO? I would advise you to install the latest version 2021.3 OpenVINO, test it out in your production environment, and share the result.

Regards,

Aznie

[IntelSupport]

Hi Yu,

I hope everything is going well for you. For your information, there is a similar issue reported to us for older OpenVINO and it should be fixed with the latest OpenVINO version.

Regards,

Aznie

[IntelSupport]

Hi Yu,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.

Regards,

Aznie