VPU HDDL plugin: RuntimeError when run example face_recognition_demo.py

[khiem_nvt]

Hello

I have installed Openvino (openvino_2021.4.752) and installed Mustang-V100-MX8 driver based on https://docs.openvino.ai/latest/openvino_docs_install_guides_installing_openvino_linux_ivad_vpu.html.

After that I verified by running scripts:
./demo_squeezenet_download_convert_run.sh -d HDDL

./demo_security_barrier_camera.sh -d HDDL

Both script worked well.

And next, I tested face_recognition_demo.py

It works well with CPU Plugin, but has error when using HDDL plugin. Log as below:
python3 face_recognition_demo.py -i videos/test_digitalsignage.mp4 -d_fd HDDL -d_lm HDDL -d_reid HDDL -m_fd intel/face-detection-retail-0004/FP16/face-detection-retail-0004.xml -m_lm intel/landmarks-regression-retail-0009/FP16/landmarks-regression-retail-0009.xml -m_reid intel/face-reidentification-retail-0095/FP16/face-reidentification-retail-0095.xml --verbose -fg "/opt/intel/openvino_2021/deployment_tools/open_model_zoo/demos/face_recognition_demo/python/gallery" --run_detector
[ INFO ] Initializing Inference Engine...
[ INFO ] Loading networks...
[ INFO ] Building faces database using images from "/opt/intel/openvino_2021/deployment_tools/open_model_zoo/demos/face_recognition_demo/python/gallery"
Traceback (most recent call last):
File "face_recognition_demo.py", line 303, in <module>
sys.exit(main() or 0)
File "face_recognition_demo.py", line 240, in main
frame_processor = FrameProcessor(args)
File "face_recognition_demo.py", line 157, in __init__
self.face_detector if args.run_detector else None, args.no_show)
File "/home/netvision/khiem/face_recognition_demo/python/faces_database.py", line 66, in __init__
rois = face_detector.infer((image,))
File "/home/netvision/khiem/face_recognition_demo/python/ie_module.py", line 69, in infer
return self.postprocess()
File "/home/netvision/khiem/face_recognition_demo/python/face_detector.py", line 87, in postprocess
outputs = self.get_outputs()[0][self.output_blob].buffer
File "/home/netvision/khiem/face_recognition_demo/python/ie_module.py", line 56, in get_outputs
self.wait()
File "/home/netvision/khiem/face_recognition_demo/python/ie_module.py", line 51, in wait
self.perf_stats[i] = self.exec_net.requests[i].get_perf_counts()
File "ie_api.pyx", line 1533, in openvino.inference_engine.ie_api.InferRequest.get_perf_counts
File "ie_api.pyx", line 1560, in openvino.inference_engine.ie_api.InferRequest.get_perf_counts
RuntimeError: [ GENERAL_ERROR ] AssertionFailed: deviceTimings != nullptr

I use Ubuntu20.04 with Linux netvision-SER0 5.13.0-27-generic #29~20.04.1-Ubuntu SMP Fri Jan 14 00:32:30 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux

Please help.

Best regards,

Khiem Phung

[Wan_Intel]

Hi Khiem_nvt,

Thanks for reaching out to us.

We encountered the same error when running the Face Recognition Python Demo using HDDL plugin.

 

We are investigating this issue and we will update you at the earliest.

Regards,

Wan

[khiem_nvt]

Hello,

Thank for confirming that bug, I hope that bug will be resolved soon 

Regards

Khiem Phung

[Wan_Intel]

Hi Khiem Phung,

Thanks for your patience.

Our developer responded they are investigating this issue, and we will update you as soon as we solve the issue.

Regards,

Wan

[khiem_nvt]

Hi,

Any new or update about this issue for last 2 weeks?

Regards,

Khiem Phung

[Hari_B_Intel]

Hi Khiem_nvt,

Thank you for your patience, our developer is looking into this issue. The issue that you are facing is from the inference API library, so it might take some time from the developer side to fix the HDDL.

Worry not, we will try our best to get a response soon.

Thank you

[imnilesh]

Hello, 

I am getting the similar issue while running face recognition demo, did you fixed this issue?

Please update.

Thanks

[Hari_B_Intel]

Hi imnilesh,

Thank for reaching out to use, for your information our developer still fixing this HDDL issue, we will get back to you once available.

Thank you

[Hari_B_Intel · Accepted solution]

Hi imnilesh, khiem_nvt

The issue seems to be resolved in 2022.1 release of OpenVINO. I've just tried the face_recognition_demo with HDDL plugin (card with 2x MyriadX VPUs) and the Python demo seems to be executing normally. 

$ python3 /opt/intel/openvino_2022.1.0.643/open_model_zoo/demos/face_recognition_demo/python/face_recognition_demo.py -i ~/media/videos/face-demographics-walking.mp4 -m_fd ~/openvino_models/2022.1/intel/face-detection-retail-0004/FP16/face-detection-retail-0004.xml -d_fd HDDL -m_reid ~/openvino_models/2022.1/intel/face-reidentification-retail-0095/FP16/face-reidentification-retail-0095.xml -d_reid HDDL -m_lm ~/openvino_models/2022.1/intel/landmarks-regression-retail-0009/FP16/landmarks-regression-retail-0009.xml -d_lm HDDL --loop

Please have a try and if you still face the issue with HDDL, feel free to post on this channel.

face_recognition_demo with HDDL

Thank you

[imnilesh]

Sure Hari, I will check that and thanks for your response.

[Wan_Intel]

Hi Khiem_nvt and Imnilesh,

Thanks for your question.

This thread will no longer be monitored since this issue has been resolved. 

If you need any additional information from Intel, please submit a new question.

Regards,

Wan