• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 30 subscribers
  • Views 1590 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.

IWR6843: Object Classification in mmWave Radar demos

Arun MP
Intellectual 615 points
Part Number: IWR6843

Hi Team,

We are working on object classification inside a room, classify human and non-human objects. As part of this I used Indoor flase detection demo and run in a room which is having a chair and human. But I haven't seen that the classification is not consistent. I observed the following,

1. Most of the time it was not classifying and giving BLACK colour bubbles in GUI.

2. Its very rare that detect both the objects and classified as human and non-human objects. For human It was showing GREEN colour and for chair it was showing GERY.

3. In some cases it was showing both the objects as GREY in colour.

So could you please guide us how the classification is done. If possible direct us to the code to find out the same.

Thanks

Arun

  • 0
    TI__Mastermind 28055 points

    Hi Arun,

    First, you can find the classifier source code here; C:\ti\mmwave_industrial_toolbox_4_0_0\labs\people_counting\68xx_indoor_false_detection_mitigation\src\common\modules\classification

    The classification demo does not detect static objects - so if a object is not moving, it should not generate detection points, and should not be tracked by the tracker. This means that you must be moving the chair to get it become detected. The demo has been trained to identify humans, plants, fans and curtains - it has not been trained with chairs so the behavior will be undefined - plus the chair should not be detected as previously mentioned. 

    I recommend trying the following configuration - replace the classifiierCfg line in your configuration with the following: 

    classifierCfg 1 1 3 500 0.8 1.0 0.95 10

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Intellectual 615 points

    Hi Justin,

    I agree that it does not detect the static object. I forgot to mention that we have used an operator chair and was rotating by a person who is standing one-hand distance.

    At that time it was showing multiple objects in GUI with different colours.

    I will follow the configuration that you provided and analyse the output.

    Thanks for reply,

    Arun