Libnvds_nvmultiobjecttracker.so failed to load REID tracker

**• Hardware Platform (Jetson / GPU): A2000
**• DeepStream Version :6.3
**• TensorRT Version: 8.6.1

When my program is trying to load any model as tracker it’s crushing.

Last messages:

[INFO] There are 6 embeddings in database
defining argumentsNumber of instances ------------ 1
** WARN: <parse_dsexample:907>: Unknown key 'batch-size' for group [ds-example]
** WARN: <parse_source:657>: Unknown key 'rtsp-reconnect-interval' for group [source0]

(watchcam-Nano-JBL_latest2:330231): GLib-GObject-WARNING **: 19:55:25.782: g_object_set_is_valid_property: object class 'GstDsExample' has no property named 'blur-objects'
Unknown or legacy key specified 'is-classifier' for group [property]
Warn: 'threshold' parameter has been deprecated. Use 'pre-cluster-threshold' instead.
befordsgstnvtracker: Loading low-level lib at /opt/nvidia/deepstream/deepstream-6.3/lib/libnvds_nvmultiobjecttracker.so
[NvMultiObjectTracker] Loading TRT Engine for tracker ReID...
19:55:26: The program has unexpectedly finished.
19:55:26: The process was ended forcefully.
19:55:26: /home/sigmind/Documents/FFL/WC-Server-DS/watchcam-Nano-JBL_latest2 crashed.

What can be the reason?

Here is my output for gst-inspect-1.0 nvinfer

inspect.txt (6.9 KB)

Which application are you running? Can you reproduce the issue with DeepStream sample application?

I am using the deepstream_app.c and the parser that has been given with the SDK. I have integrated them in a single .cpp file and using it load the models. Actually I intent to collect embedding as it is shown in deepstream_app.c. I am using a arcface model as tracker. In the stand alone deepstream application the model loads fine as REID tracker. The model loads fine as sgie in my Qt application and I can collect embedding. But when I try to use it as a tracker it fails. I can’t reproduce the issue in the sample application as it loads fine. Can I produce some error messages to check why the TRT engines fails to load the model? Also where can I get the source code for libnvds_nvmultiobjecttracker.so?

We haven’t release the source code of libnvds_nvmultiobjecttracker.so. Can you refer this guide for setup Re-ID model?: Gst-nvtracker — DeepStream documentation 6.4 documentation

I have setup the REID tracker exactly the way you are referring to. In deepstream_app.c the config file is working fine. Problem occurs in my Qt application implementation. If you could refer me to some TensorRT debugging utilities it would be helpful. If there is a way to know exactly the reason why the model fails to load.

Here is the guide for customize Re-ID model. Does your model we can download on public network? Can you share the input/output and mode details on your model? : Gst-nvtracker — DeepStream documentation 6.4 documentation (nvidia.com)

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

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