Using Pre-trained models in NCS2

[Kaliyappan__Malathi]

Hi,

I need a clarification about Neural Compute Stick 2 using pre-trained models.

Is it possible to use pedestrian detection and action recognition (models build in OpenVINO R3 version) in Intel neural compute stick 2 ?

Thanks in advance.

[Luis_at_Intel]

Hi Kaliyappan, Malathi

Please take a look at this document as it has a list of the Pre-Trained models, supported demos and plugins. According to the document for pedestrian detection, action recognition models and Myriad plugin (NCS2) these are supported:

• pedestrian-detection-adas-0002
• person-detection-action-recognition-0005

Regards,

Luis

[Kaliyappan__Malathi]

Hi Luis,

Thanks for response.

I need a clarification about deploying my model in NCS2

1) My OpenVINO action recognition model (Pre-trained) using IP camera for taking input in RTSP format (rtsp://admin:admin@192.168.1.108:554/cam/realmonitor?channel=1&subtype=0). Can please let me know If I can access the IP camera from NCS2, after I deploy the model on that.

Thanks

Malathi

[Luis_at_Intel · Accepted solution]

Hi Malathi,

You should be able to use the frames from IP camera with your NCS2. I found some threads from other users tying to do something similar with IP cameras and OpenVINO:

• Access RTSP stream with OpenCV 4.1.0-openvino
• how to use IP cam with openvino

Hope this helps, let me know if you have additional questions.

Regards,

Luis