NvDsAnalytics plugin line counting accuracy

[bbjor1]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson Xavier 32
• DeepStream Version 6
• JetPack Version (valid for Jetson only) 4.6
• TensorRT Version 8.5
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs)
• 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)

Hi,

We are using the NvDsAnalytics plugin to count people as they enter and exit a building. Our detector is quite good, and specifically trained for this purpose using the CrowdHuman dataset.

One issue we are having is that the line counting plugin accuracy is very low, our error rate is 18% compared to counting manually. We are using “strict” mode, and “extended=0”. Any ways to make this more accurate?

Best regards,
Brede

[fanzh]

could you share the source video and configuration file?

[bbjor1]

I can’t share source video as its from a customer location.

Analytics config looks like this:

[line-crossing-stream-0]
enable=1
line-crossing-Entry=241;654;254;589;2;566;454;570;
line-crossing-Exit=270;523;282;477;128;453;412;443;
class-id=0
extended=0
mode=strict

We’ve tried using different placements of the lines, there is some improvements in accuracy, but the overall impression is that its not accurate enough.

[fanzh]

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

the effect of video analysis is related to detection’s bboxes and crossing line, can you use deepstream native video and detection model to reproduce this issue?