yolov2 issue on Python sample

[Chakraborty__Subhasi]

I am trying to use yolov2 on openvino R3 through python. I am getting the following error: RuntimeError: 

"Unsupported primitive of type: RegionYolo name: output/YoloRegion/teamcity/work/scoring_engine_build/releases_openvino-2018-r3/ie_bridges/python/inference_engine/ie_api_impl.cpp:226"

What do i do next to run my code successfully? I have heard that Region Layer is already supported in Inference Engine of this new openvino version

[nikos1]

Are you trying on CPU or GPU?

I have a potentially similar issue here with Python tiny yolo v2 only when on CPU  When using GPU device yolo runs fine. Here is my error

[ INFO ] Initializing plugin for CPU device...
[ INFO ] Reading IR...
[ ERROR ] Following layers are not supported by the plugin for specified device CPU:
 output/YoloRegion
[ ERROR ] Please try to specify cpu extensions library path in sample's command line parameters using -l or --cpu_extension command line argument

[Severine_H_Intel]

Dear Subhasis and Nikos,

try to add the argument -l <path_to_samples_build>\intel64\Release\cpu_extension.dll (\lib\libcpu_extension.so for Linux). Yolo includes some layers not known to MKLDNN directly and this is why we provide this library extension. The clDNN (behind the GPU plugin) knows them, explaining it already works for GPU without adding this path. 

This is specific to the Python samples, as the C++ sample loads those libraries automatically. 

Best, 

Severine

[Chakraborty__Subhasi]

Thank you Nikos and Severine for your insight. After changing it from CPU to GPU, the code runs succesfully.

But there are no detections. I cannot figure out the problem. Could you please throw some light.

[nikos1]

Hi Subhasis, may be better to create a new ticket for the detection and how to process the YOLO output issue so that we can better focus on the RuntimeError on CPU you reported originally.

[nikos1]

Hi Severine,

Thank you for your help. I found the following libraries that seem to get me further down in the pipeline. Will do some more testing and report back here.  

computer_vision_sdk_2018.3.343/deployment_tools/inference_engine/lib/ubuntu_16.04/intel64/libcpu_extension_avx2.so
computer_vision_sdk_2018.3.343/deployment_tools/inference_engine/lib/ubuntu_16.04/intel64/libcpu_extension_sse4.so

[lai__autumn]

您好，我使用openvino中sample.收到錯誤：RuntimeError:
python3 object_detection_demo_yolov3_async.py 
--labels /home/autumn/tensorflow-yolo-v3-master/obj.names 
-i /home/autumn/tensorflow-yolo-v3-master/222.mp4 
-m /home/autumn/tensorflow-yolo-v3-master/FP32/frozen_darknet_yolov3-tiny_final.xml 
-d CPU 
--cpu_extension /opt/intel/openvino/deployment_tools/inference_engine/lib/intel64/libcpu_extension_sse4.so

得到結果如下：
 INFO ] Loading network files:
    /home/autumn/tensorflow-yolo-v3-master/FP32/frozen_darknet_yolov3-tiny_final.xml
    /home/autumn/tensorflow-yolo-v3-master/FP32/frozen_darknet_yolov3-tiny_final.bin
Traceback (most recent call last):
  File "object_detection_demo_yolov3_async.py", line 349, in <module>
    sys.exit(main() or 0)
  File "object_detection_demo_yolov3_async.py", line 189, in main
    assert len(net.outputs) == 3, "Sample supports only YOLO V3 based triple output topologies"
AssertionError: Sample supports only YOLO V3 based triple output topologies

請問，要如何解決？謝謝！！