Deepstream tracker fail to assign same id to same people

[saurav.singh]

Deepstream tracker fail to assign same id to same people. I following all the instructions provided with Nvidia and I have also tried different trackers but I am still getting the same results.

For more detail I am using.

1. Deepstream 6.1
2. peoplenet_deployable_quantized_v2.6
3. NVIDIA Tesla V100

For reference

I have used deepstream:6.1-devel image and for the configuration I have used
deepstream_app_source1_classifier_models.yml which you can find using this link deepstream_reference_apps/deepstream_app_tao_configs at master · NVIDIA-AI-IOT/deepstream_reference_apps · GitHub and the output file you can find using this link out.mp4 - Google Drive for tracker I have tried with

config_tracker_NvDCF_perf.yml
config_tracker_NvDCF_accuracy.yml
config_tracker_DeepSORT.yml
config_tracker_IOU.yml

[kesong]

Yes, I can see the ID switch. But all kind of tracker have this kind of issue. We can’t totally fix the ID switch.

[saurav.singh]

Hi, Thanks for the reply. I understood the problem but if I am building an attendance system which will count the number of people in a room so in that case what can I do. Is there any parameters which I can change which will give me better accuracy.

[kesong]

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

Can you have a try with nvdsanalytics?

https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvdsanalytics.html