• State Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 4 answers
  • Subscribers 31 subscribers
  • Views 519 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

IWR6843AOPEVM: Next Steps for tuning for clean accurate people count

Sagar Jaiswal
Prodigy 230 points
Part Number: IWR6843AOPEVM
Other Parts Discussed in Thread: IWR6843

Hello Everyone,

Apologies if this is a repost question but we are still experimenting with this technology.

We managed to implement the people counting demo as well as run the visualizer. Couple question that arose:

-Is it possible to tune the sensor remotely? (for a specific room doing it on the spot instead of hard coding for the room)

-is there OTA available for the sensor (over the air update)?

-What are the necessary next steps for tuning? What exactly should be focused on to get a clean accurate people count?

-How to extract the data from the sensor to a raspberry pi/or any other computer that support serial/linux?

Apologies if these questions are broad/vague but any help will really be appreciated!

Kind Regards

Sagar Jaiswal

  • 0
    TI__Mastermind 28055 points

    Hi Sagar,

    1. We have a customization guide that explains different knobs that are available. Typically you want to use CFAR to modify the point cloud so that you as a human can tell where people are when looking at the device output, then tune the tracker to work well with that point cloud.
    2. Yes, see the Secondary Boot loader.
    3. See #1
    4. The data format of output is described in the user's guide. It is on a UART connection at 921600 Baud.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 230 points

    Hello

    Thank you for your quick  reply.

    Could you please provide link to the customization guide? And maybe the link to this CFAR you are talking about please?

    As for this secondary boot loader, how would this work over wifi? Since the IWR6843 doesn't have Wifi.

    Also do you have any links to examples for the UART connection or any serial connections?

    EDIT:

    Sorry just wanted to ask a followup question, to work as a peope counting product is there any major changes required to the code or would just tuning the parameters suffice?

    Thanks

    Kind Regards

    Sagar Jaiswal

  • 0 in reply to Sagar Jaiswal
    TI__Mastermind 28055 points

    Hi Sagar,

    You can find the customization guide inside the docs folder of the 3D People Counting lab once you have downloaded it from TIREX.

    CFAR is described here: https://en.wikipedia.org/wiki/Constant_false_alarm_rate You can see how to tune it in the People Counting lab by looking at the document "3DPeoplecountingDemo_ConfigurationDetails"

    You have to use a wifi device to interact with the IWR6843. 

    I am not sure what you are looking for with a UART example. Every lab uses UART to enable the CLI and send data out of the device.

    For going to production, you will have to determine what your requirements are , and how far from those requirements the demo code is.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 230 points

    Awesome, thank you .

    Thank you for the resources, as for the UART, i think i figured it out.

    Just one last question, in terms of production, there are no specific requirements, its literally just detecting occupancy in a room or a hallway, or detecting the number of people in a line. So for this, correct me if I'm wrong, I most likely do not need to change anything with the algorithms or in the code, and would most likely just need to change the parameters and such, yes?

    Thanks for the help!

    Kind Regards

    Sagar

  • 0 in reply to Sagar Jaiswal
    TI__Mastermind 28055 points

    Hi Sagar,

    The algorithm can be tuned to a point - and is good for detecting people in a room. Our most production ready demo is the "Sense and Avoid: HVAC Control". However, my comment is these demos are built first and foremost to showcase capabilities of the device and give user's code examples to work with. These aren't fine tuned to a specific use case, so I would expect that some combination of tuning or code modification will be required to maximize performance in your specific use case.

    Regards,

    Justin 

  • 0 in reply to Justin Curewitz
    Prodigy 230 points

    Justin Curewitz said:

    Hi Sagar,

    The algorithm can be tuned to a point - and is good for detecting people in a room. Our most production ready demo is the "Sense and Avoid: HVAC Control". However, my comment is these demos are built first and foremost to showcase capabilities of the device and give user's code examples to work with. These aren't fine tuned to a specific use case, so I would expect that some combination of tuning or code modification will be required to maximize performance in your specific use case.

    Regards,

    Justin 

    Great, I understand .

    Well in that case, I have been looking at the code and it's sort of difficult to understand the flow of the firmware. Is there maybe a guide or something that explains the code or something I can read up? If so, could you please provide the link?

    For example, it has two projects mss and dss, I can compile them but I do not fully understand the entry, exit points and what I would need to change specifically and how the two mss and dss link together? 

    Thank you, you've been a great help so far!

    Kind Regards

    Sagar Jaiswal

  • 0 in reply to Justin Curewitz
    Prodigy 230 points

    Hello  

    Thank you for the links. Final question, is there a link to the SDK API? Where it could explain the functions and provide a description of every function available for the mmWAVE?

    Thank you!

    Kind Regards

    Sagar

  • 0 in reply to Sagar Jaiswal
    TI__Mastermind 28055 points

    Hi Sagar,

    After you download the mmWave SDK - open <mmwave_sdk_install_folder>/docs/mmwave_sdk_module_documentation.html

    Regards,

    Justin