Arbitrary large object IDs with libnvds_nvdcf!

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1
**• Hardware Platform--------Jetson nano**
**• DeepStream Version-------5.0**
**• JetPack Version -----------4.4.1**
**• TensorRT Version----------7.1.3-1+cuda10.2**

gst-launch-1.0 uridecodebin3 uri="file:///opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.mp4" ! videorate drop-only=true max-rate=10 ! mux.sink_0 nvstreammux live-source=0 name=mux batch-size=1 batched-push-timeout=40000 width=1280 height=720 ! queue ! nvinfer config-file-path=/opt/nvidia/deepstream/deepstream/sources/apps/sample_apps/deepstream-test1/dstest1_pgie_config.txt unique-id=1 ! nvtracker ll-lib-file=/opt/nvidia/deepstream/deepstream/lib/libnvds_nvdcf.so tracker-width=1280 tracker-height=704 ! nvvideoconvert ! nvdsosd ! nvegltransform ! nveglglessink -v

I am observing very large object IDs with libnvds_nvdcf. With other two default trackers, IDs are normal.

How can this issue be resolved?

Thanks

Screenshot from 2020-12-03 23-23-57
Screenshot from 2020-12-03 23-23-571300×766 148 KB

3

Yes, this is a problem of current DCF trackID, only low 32bit are valid for the 64bit parameter. But the value is unique for every tracked object, so it will not impact anything.

4

Thank you.

Inside osd_sink_pad_buffer_probe,
Can obj_meta->object_id be modified to correct IDs in display?

I tried but display is still showing same problem.

.
obj_meta = (NvDsObjectMeta *)(l_obj->data);
.
guint32 *id = (guint32 *)&obj_meta->object_id;
obj_meta->object_id = id[0]; // I have verified id[0] is correct ID

I am trying this in a C++ program, trying to simulate above pipeline.
Reference for me is the ‘deepstream_app.c’

5

You can try that.

6

Sorry, I tried this, but it is not working, what is the proper way to modify osd data?

7

Can you try to set “useUniqueID” in your DCF tracker config file to 0(the default value is 1)? Gst-nvtracker — DeepStream 6.1.1 Release documentation