I want to run the smart_classroom_demo on windows 10 with openvino_2020.2.117, using the command:
smart_classroom_demo.exe -i "cam" -m_act "person-detection-action-recognition-0003.xml" -m_fd "face-detection-adas-0001.xml" -m_lm "landmarks-regression-retail-0009.xml" -m_reid "face-reidentification-retail-0095.xml"
but I get the error:
MKLDNNGraph::CreateGraph: such topology cannot be compiled for dynamic batch!
However, it works when I changed device to GPU.
There are two possible reasons for being able to run the demo in GPU, but not in CPU.
Firstly, OpenVINO supports only 6th to 10th generation Intel® Core™ processors. Please check your CPU details to ensure that it conforms to OpenVINO requirements.
The relevant information regarding these can be found at the following link:
The second possibility is that the pre-trained models that you are using contain layers that are not fully supported by CPU for Dynamic Batching feature. There are layers where CPU support is implemented via Extensibility Mechanism, whilst for some layers, CPU supports NCDHW layout.
The list of layers supported by the different plugins is available at the following page:
Meanwhile, the list of layers supported for Dynamic Batching is available here:
The following article will walk you through setting up and running the Smart Classroom Demo on Windows.