<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic Re:NCS2 Parallel Networks on one Device are Serialized? in Intel® Distribution of OpenVINO™ Toolkit</title>
    <link>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1213928#M20893</link>
    <description>&lt;P&gt;Hi Dominik,&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;&lt;/P&gt;&lt;P&gt;Do you need any other additional information?&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;&lt;/P&gt;&lt;P&gt;Regards,&lt;/P&gt;&lt;P&gt;Rizal&lt;/P&gt;&lt;BR /&gt;</description>
    <pubDate>Thu, 01 Oct 2020 01:55:13 GMT</pubDate>
    <dc:creator>Rizal_Intel</dc:creator>
    <dc:date>2020-10-01T01:55:13Z</dc:date>
    <item>
      <title>NCS2 Parallel Networks on one Device are Serialized?</title>
      <link>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1211092#M20748</link>
      <description>&lt;P&gt;Hi everyone,&lt;/P&gt;
&lt;P&gt;I successfully created a network that I can infer on the NCS2. Now I want to speed up the inference by using 5 networks in parallel on 1 NCS2 where each network should use 4 inference requests.&lt;/P&gt;
&lt;P&gt;I do this the following way:&lt;/P&gt;
&lt;OL&gt;
&lt;LI&gt;I spawn 5 threads where each creates an InferenceEngine::ExecutableNetwork for the MYRIAD device This should load 5 networks onto the NCS2.&lt;/LI&gt;
&lt;LI&gt;In each of those threads I spawn 4 additional threads that create a synchronous inference request.&lt;BR /&gt;&lt;BR /&gt;&lt;/LI&gt;
&lt;/OL&gt;
&lt;P&gt;Everything runs without errors and also the output image is correct but I don't see any speedup. When I measure the computation time it's almost the same (140seconds +- 1s) as when I run only one network. It looks like only one network and one inference request is doing all the work.&lt;BR /&gt;&lt;BR /&gt;Does anyone have similar issues. Maybe there is something wrong with my architecture?&lt;BR /&gt;&lt;BR /&gt;Best Regards&lt;/P&gt;
&lt;P&gt;Dominik&lt;/P&gt;</description>
      <pubDate>Mon, 21 Sep 2020 09:12:15 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1211092#M20748</guid>
      <dc:creator>Dominik</dc:creator>
      <dc:date>2020-09-21T09:12:15Z</dc:date>
    </item>
    <item>
      <title>Re:NCS2 Parallel Networks on one Device are Serialized?</title>
      <link>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1211450#M20765</link>
      <description>&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;Hi Dominik,&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;There seems to be no problem with your architecture based on your explanation.&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;There is only this &lt;/SPAN&gt;&lt;A href="https://docs.openvinotoolkit.org/latest/openvino_docs_optimization_guide_dldt_optimization_guide.html#myriad" rel="noopener noreferrer" target="_blank" style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;requirement&lt;/A&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt; to parallelize the workload as much as possible.&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;Actually, there is only a single Movidius X chip on the NCS2. Therefore, inference calls are actually queued to be executed on the single chip.&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;You could check your parallel implementation with &lt;/SPAN&gt;&lt;A href="https://github.com/openvinotoolkit/open_model_zoo/tree/2018/demos/crossroad_camera_demo" rel="noopener noreferrer" target="_blank" style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;crossroad camera demo&lt;/A&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt; and &lt;/SPAN&gt;&lt;A href="https://github.com/openvinotoolkit/open_model_zoo/tree/master/demos/python_demos/action_recognition" rel="noopener noreferrer" target="_blank" style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;action recognition demo&lt;/A&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;.&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;To get an increase in performance you would need to have multiple NCS2 sticks (or &lt;/SPAN&gt;&lt;A href="https://software.intel.com/content/www/us/en/develop/topics/iot/hardware/vision-accelerator-movidius-vpu.html" rel="noopener noreferrer" target="_blank" style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;Intel® Vision Accelerator Design&lt;/A&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;). There is an &lt;/SPAN&gt;&lt;A href="https://github.com/yuanyuanli85/open_model_zoo/blob/ncs2/demos/python_demos/multiple_device_ncs2_async.py" rel="noopener noreferrer" target="_blank" style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;example&lt;/A&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt; you can use a reference created by Victor Li for utilising multiple NCS2. This is based on the concept of &lt;/SPAN&gt;&lt;A href="https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_supported_plugins_MYRIAD.html#MYRIAD_DEVICE_ALLOC" rel="noopener noreferrer" target="_blank" style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;MYRIAD device allocation&lt;/A&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;.&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;Regards,&lt;/SPAN&gt;&lt;/P&gt;&lt;P&gt;&lt;SPAN style="font-family: Calibri, sans-serif; font-size: 11pt;"&gt;Rizal&lt;/SPAN&gt;&lt;/P&gt;&lt;BR /&gt;</description>
      <pubDate>Tue, 22 Sep 2020 11:32:36 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1211450#M20765</guid>
      <dc:creator>Rizal_Intel</dc:creator>
      <dc:date>2020-09-22T11:32:36Z</dc:date>
    </item>
    <item>
      <title>Re: Re:NCS2 Parallel Networks on one Device are Serialized?</title>
      <link>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1212545#M20821</link>
      <description>&lt;P&gt;Thank you for you feedback. The time difference was just so little that I didn't recognize it at first but now it works fine.&lt;BR /&gt;&lt;BR /&gt;BR&lt;BR /&gt;Dominik&lt;/P&gt;</description>
      <pubDate>Fri, 25 Sep 2020 16:52:56 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1212545#M20821</guid>
      <dc:creator>Dominik</dc:creator>
      <dc:date>2020-09-25T16:52:56Z</dc:date>
    </item>
    <item>
      <title>Re:NCS2 Parallel Networks on one Device are Serialized?</title>
      <link>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1213928#M20893</link>
      <description>&lt;P&gt;Hi Dominik,&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;&lt;/P&gt;&lt;P&gt;Do you need any other additional information?&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;&lt;/P&gt;&lt;P&gt;Regards,&lt;/P&gt;&lt;P&gt;Rizal&lt;/P&gt;&lt;BR /&gt;</description>
      <pubDate>Thu, 01 Oct 2020 01:55:13 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1213928#M20893</guid>
      <dc:creator>Rizal_Intel</dc:creator>
      <dc:date>2020-10-01T01:55:13Z</dc:date>
    </item>
    <item>
      <title>Re:NCS2 Parallel Networks on one Device are Serialized?</title>
      <link>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1215095#M20936</link>
      <description>&lt;P&gt;Hi Dominik,&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;&lt;/P&gt;&lt;P&gt;Intel will no longer monitor this thread since we have provided a solution.&amp;nbsp;If you need any additional information from Intel, please submit a new question.&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;&lt;/P&gt;&lt;P&gt;Regards,&lt;/P&gt;&lt;P&gt;Rizal&lt;/P&gt;&lt;BR /&gt;</description>
      <pubDate>Tue, 06 Oct 2020 02:39:26 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Distribution-of-OpenVINO/NCS2-Parallel-Networks-on-one-Device-are-Serialized/m-p/1215095#M20936</guid>
      <dc:creator>Rizal_Intel</dc:creator>
      <dc:date>2020-10-06T02:39:26Z</dc:date>
    </item>
  </channel>
</rss>

