openvino 2020.3 demo : multi_channel_face_detection_demo , GPU + yolov3 not work ..?

[Yao__Kent]

1. Environment: Ubuntu 18.04 and demo app from Openvino 2020.3

2. After running multi_channel_face_detection_demo -d GPU yolov3 model  , the output shows   => "Can't get ngraph::Function. Make sure the provided model is in IR version 10 or greater" . But running multi_channel_face_detection_demo -d HDDL or  -d CPU  with the same yolov3 model , it can work .

3. Another app : interactive_face_detection_demo -d GPU with face-detection-retail-0004 mode , it can work, (i think the GPU device  may work normally... )

4. I build the yolov3 model by using parameters below:

mo.py --data_type=FP16 --output_dir.\\FP16\\--input_model.\\Generic_Models\tensorflow-yolo-v3\frozen_darknet_yolov3_model.pb --batch 1 --tensorflow_use_custom_operations_config.\\Generic_Models\tensorflow-yolo-v3\yolo_v3.json 

what is wrong with yolov3 model ?? or other reason?

Thanks

[Yao__Kent]

Sorry,

The title update: multi_channel_object_detection_demo_yolov3

[Munesh_Intel]

Hi Kent,

Greetings to you.

We are currently investigating this issue, and will get back to you at the earliest.

Regards,

Munesh

[Max_L_Intel · Accepted solution]

Hi @Yao__Kent 

There's a known issue with object_detection_demo_yolov3_async, and seems it is supposed to be the same ngraph function error message for multi_channel_object_detection_demo_yolov3 too.
The fix for the first one should be implemented in the next software release, and the hot fix has been already posted here https://github.com/eizamaliev/open_model_zoo/commit/929d673e1176ce99f8d49d1e86ff31f2a4f25c74 

With regards to Multi Channel version of Object Detection yolov3 demo, we have shared this information with developers team.

Thanks for reporting this.

[Max_L_Intel]

This should already be fixed in OpenVINO 2020.4 release.