OpenVINO Pre-trained model support on Movidius

[Yalee]

Hello,

Does Moviduis support ALL listed pre-trained model from OpenVINO? specifically faster-rcnn-resnet101-coco-sparse-60-0001 with 50+ MParam. I'm trying to get an idea what pre-trained model from OpenVINO can and cannot support on Movidius and its performance (FPS). thanks

[Syamimi_Intel · Accepted solution]

Hi Aaron Lee,

Thank you for reaching out.

The Inference Engine supports inference of deep learning models using Movidius products via VPU plugins, namely MYRIAD plugin and HDDL plugin.

 MYRIAD Plugin - Supports Intel Neural Compute Stick 2 powered by the Intel Movidius Myriad X

 HDDL Plugin - Supports Intel Vision Accelerator Design with Intel Movidius VPUs

For your information, with OpenVINO 2020.4 release, Intel Movidius Neural Compute Stick powered by the Intel Movidius Myriad 2 is no longer supported. Therefore, if you want to use Intel Movidius Neural Compute Stick, you have to use OpenVINO 2020.3 version.

You can refer to the following link to find the supported networks for MYRIAD plugin and HDDL plugin. As you can see, faster-rcnn-resnet101 is listed under the TensorFlow subsection, which means that it's supported: https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_supported_plugins_MYRIAD.html#supported_networks

For performance measurements, you can refer to the link below, which contains benchmarked performance values for selected models.

https://docs.openvinotoolkit.org/2021.3/openvino_docs_performance_benchmarks_openvino.html

Regards,

Syamimi Fauzi

[Syamimi_Intel]

Thank you for your question. If you need any additional information from Intel, please submit a new question as this thread is no longer being monitored.