Line crossing mode and obj-cnt-win-in-ms parameter

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.3
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs) question/bug
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

Hello there,
I have been using ds_analytics with ROI and Line Crossing for a while, but started noticing strange behaviour.

  1. Line crossing is detected when the object appears very far from the line. At no point is the bottom center of the bbox intersecting with the line. (Can’t share the actual footage but the picture illustrates the situation quite well)

line_crossing_problem_20270607
line_crossing_problem_202706071152×648 4.17 KB

  1. It appears that line crossing counts objects moving in the opposite direction with respect to the direction arrow.

Extended is set to 0 and mode is set to loose.

So question 1 is - how should we understand the meaning of loose? Would it be able to “catch” the objects moving in direction 180 degrees different than defined direction?

As for debugging my first problem - I’ve seen this post False line-crossing count for persons who do not pass the line in deepstream-occupancy-analytics but it relates to DS6.0 and Jetson. Should I consider applying the same patch that is mentioned there to DS6.3 on dGPU?

And what exactly is the meaning of obj-cnt-win-in-ms=2 parameter and how should one tune it? I see it in the source code but don’t see it documented anywhere in the developer guide.
Is it even relevant in my case as the begginning when the object is tracked is far from the line so it should not be affected by the initial detections of this object.

I currently use interval of 2 for inference so my final throughput is around 6FPS.

4

please help to help to reproduce this issue. could you share the configuration file, code diff and video by the forum private email. please click forum avatar-> personal messages->new message.

you can find the loose explanation in the doc.

In that case, the fps is only 2. this patch is only for very low fps case. as the code comment in gstnvdsanalytics.h shown, it means “Window in ms for obj count”, which is used for low-level strategy.

5

I’m sorry, but unfortunately we cannot share the video used in this case.

I’ve seent the definition for loose/balanced/strict that you linked to, but I was hoping that there exists a more scientific answer.

But I have two other questions:

  • If an identified object, for which the tracking id remains stable, crosses the line multiple times and from the right direction - would it be counted as line crossing multiple times or just once?

  • Have there been any changes or improvements in DS6.4 or DS7.0 in the analytics plugin?

6
  1. if object did not change and cross the line from the right direction multiple times, it will be counted as line crossing multiple times.
  2. in the analytics plugin, there are almost the same.
7

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.