We have video analytics based application and we would like to explore possible paths in order to integrate with Milestone VMS?

[0053X00000ECHmeQAH (Amir Ismail)]

We are developing video analytics applications including but not limited to people/vehicule detection, counting, tracking. Currently, all this stuff is running on NVIDIA Jetson board and we are using your available VPService to integrate with Milestone VMS. Since the hardware ressources are a bit limited with the Jeston family board, we are thinking of switching to dGPU-based platforms.

So, our ultimate goal is to run our applications on Windows OS (together with Milestone products). That being said, we would like to have more information about the best integration choices. We have explored MIP SDK and we need your assistance to identify the type of integration that suits our case.

Thanks for your collaboration.

[Peter_Posselt_Vergmann (Peter Posselt Vergmann (Milestone Systems))]

Hi Amir,

As I see it you have multiple options:

• Our new AI Bridge framework might be of interest to you. Have a look at the presentation from our last Developer Conference and let us know if it is of interest to you: https://www.youtube.com/watch?v=cyq1Cdv31F0&t=1s
• VPS can also be built for Windows so if you want to run on a Windows platform you can still use your existing code (more or less)
• Last option if you still want to do a Windows based MIP SDK solution you can use RawLiveSource (https://doc.developer.milestonesys.com/html/index.html?base=miphelp/class_video_o_s_1_1_platform_1_1_live_1_1_raw_live_source.html&tree=tree_search.html?search=rawlivesource) or BitmapLiveSource (https://doc.developer.milestonesys.com/html/index.html?base=miphelp/class_video_o_s_1_1_platform_1_1_live_1_1_bitmap_live_source.html&tree=tree_search.html?search=bitmaplivesource) to access the video you want to analyze and then stream back either metadata using MetadataProviderService (https://doc.developer.milestonesys.com/html/index.html?base=miphelp/class_video_o_s_1_1_platform_1_1_data_1_1_media_provider_service.html&tree=tree_search.html?search=metadataprovider) or analytics events (https://doc.developer.milestonesys.com/html/index.html?base=miphelp/class_video_o_s_1_1_platform_1_1_data_1_1_analytics_event.html&tree=tree_search.html?search=analyticsevent). Some of the options are also described in this (unfortunately a bit old) introduction: https://doc.developer.milestonesys.com/html/index.html?base=gettingstarted/intro_analytics.html&tree=tree_4.html

[0053X00000FQB0jQAH (Amira Bouaouina)]

Hi Mr.Vergmann,

Thank you for your response.

First, when using the VPS on windows, we will definitely not apply the DeepStream SDK since it’s only supported by a Linux environment. So what is its alternative on windows?

Concerning the last option, after accessing the video streams that we want to analyze. We wanted to know how we could analyze them and how to deploy and apply our computer vision models.

Thank you!

[Peter_Posselt_Vergmann (Peter Posselt Vergmann (Milestone Systems))]

I am unfortunately not an expert on how to do video analytics nor which alternatives to the DeepStream SDK exists for Windows.

As to your latter question it is a bit unclear to me. Using our APIs you can get access to the raw frames of the video stream (or alternatively ones decoded into JPEG or Bitmap if desired). How you then utilize them in your software I can unfortunately not provide much input on.

[0053X00000FQB0jQAH (Amira Bouaouina)]

Thank you for your reply Mr.Vergmann and I hope you keep us in mind if you find any beneficial resources for our use case.

[HansK (Hans Kierulff (Milestone Systems))]

Hi,

Maybe Intel OpenVINO could be an option for you? It can be hosted in the VPS Toolkit gStreamer pipeline, also on Windows.

[0053X00000FQB0jQAH (Amira Bouaouina)]

Hello Mr.Kierulff,

Thank you for sharing this information with us, and we will definitely try this approach and get back to you with our results and feedback.