YOLOv3 tutorial doesn't work in latest OpenVINO, but does work in earlier version.

[Brown_Kramer__Joshua]

I recently installed openvino_2021.2.185.  I trained a yolov3 model with darknet, and I followed these instructions to convert it to openvino:

https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_YOLO_From_Tensorflow.html

I then tried running it in C:\Program Files (x86)\IntelSWTools\openvino_2021.2.185\inference_engine\demos\python_demos\object_detection_demo

Similar to another recent problem I wrote up, openvino starts loading the model onto the device, and then simply stops.  In python, no exception is thrown or anything.  The script simply stops running.  In my previous post, when I tried running from the c++ api, I got a memory access violation error during the model loading process.

I went through the same process with the same .pb file except with openvino_2020.4.287, and using c:\Program Files (x86)\IntelSWTools\openvino_2020.4.287\deployment_tools\inference_engine\demos\python_demos\object_detection_demo_yolov3_async

and it works just fine.

[Iffa_Intel]

Greetings,

You could try this solution that applies to both yolov3 and yolov4: https://github.com/TNTWEN/OpenVINO-YOLOV4/issues/1

If the issue persists, you could try to use the multi_channel_object_detection_demo_yolov3 cpp demo.

You may refer here: https://docs.openvinotoolkit.org/latest/omz_demos_multi_channel_object_detection_demo_yolov3_README.html

You need to build it first before using it.

Sincerely,

Iffa

[Iffa_Intel]

Greetings,

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

Sincerely,

Iffa