Customizing object coordinate for ROI detection in NvDsAnalytics

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

• Hardware Platform: GPU (NVIDIA L40S)
• DeepStream Version: 7.1
• CUDA Version: 12.6
• NVIDIA GPU Driver Version: 535.183.06
• Issue Type: Feature request / customization

Hello,
According to the official DeepStream documentation for the nvdsanalytics plugin:

“For all the analytics calculations, the bottom center coordinate of the bounding box of an object is being used. If the bounding box is defined as (x_left, y_top, width, height), then the bottom center coordinate would be (x_left + width/2, y_top + height).”

This behavior applies to all analytics features — including ROI-based detections (like region entry/exit, counting, etc.).

In our use case, we’d like the ability to customize the point used for these calculations. Specifically, we’d like to define a relative point in the bounding box.

My questions:

  1. Is it currently possible to configure this directly within the nvdsanalytics plugin?
  2. If not, is the source code available so that this behavior could be modified manually?

Thanks,
Yassin

3
  1. No, currently nvdsanalytics does not support customize the bottom center coordinate.
  2. nvdsanalytics plugin are opensource while the low-level implement is closedsource. you may replace the low level lib with your own algorithm lib.