The track id and the ID displayed in the display are different

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

I would like to get the tracker ID from pyds.NvDsObjectMeta.
So I tried to get the following.

However, the results obtained were different from the expected values.
The tracker ID and obj_meta.object_id displayed in the display
does not match
Why?

                # Casting l_obj.data to pyds.NvDsObjectMeta
                obj_meta=pyds.NvDsObjectMeta.cast(l_obj.data)
                obj_meta.text_params.display_text = "Person:{:.2f}".format(obj_meta.confidence)
              
                print(f"ID:{obj_meta.object_id} class id:{obj_meta.class_id} conf:{obj_meta.confidence} unique_component_id:{obj_meta.unique_component_id}")
'''
3

Hi,
Could you specify how different between print and shown on display? is it like shown on display looks like normal numbers, but print is one very long number?

4

I’ll show you.
Character string on display: person 15907525113000493130
This value is different each time depending on the person detected
Result of print: Detection ID:18446744073709551615 obj_label:person

print(f "Detection ID:{obj_meta.object_id} obj_label:{obj_meta.obj_label}")
```
5

Result of print: Detection ID:18446744073709551615 obj_label:person
This will output the same value each time.

6

We found a bug in DeepStream Python Bindings.
I can get object ID on jetson nano, but on ubuntu desktop(RTX2060), I can’t get object ID correctly.

7

This is some frame metadata about class id and tracking id,
you can see it’s not same value each time.
I am using test2 sample, may I know how can I repro your issue?

class id: 0, track id: 18105318427357872513d
class id: 0, track id: 18105318427357872521d
class id: 0, track id: 18105318427357872511d
class id: 0, track id: 18105318427357872512d
class id: 2, track id: 18105318427357872388d
class id: 2, track id: 18105318427357872471d
Frame Number = 773 Number of objects = 6 Vehicle Count = 4 Person Count = 2
class id: 0, track id: 18105318427357872522d
class id: 0, track id: 18105318427357872521d
class id: 0, track id: 18105318427357872511d
class id: 0, track id: 18105318427357872512d
class id: 2, track id: 18105318427357872388d
class id: 2, track id: 18105318427357872471d
Frame Number = 774 Number of objects = 6 Vehicle Count = 4 Person Count = 2
class id: 0, track id: 18105318427357872522d
class id: 0, track id: 18105318427357872513d
class id: 0, track id: 18105318427357872511d
class id: 0, track id: 18105318427357872512d
class id: 2, track id: 18105318427357872471d
class id: 2, track id: 18105318427357872388d
Frame Number = 775 Number of objects = 6 Vehicle Count = 4 Person Count = 2
class id: 0, track id: 18105318427357872513d
class id: 0, track id: 18105318427357872511d
class id: 0, track id: 18105318427357872512d
class id: 2, track id: 18105318427357872388d
class id: 2, track id: 18105318427357872471d
Frame Number = 776 Number of objects = 5 Vehicle Count = 3 Person Count = 2

8

Hi natukage405,

Have you managed to get issue resolved? Any result can be shared?

Thanks

10

Hi,
I have a very similar question. Is there a way to get the tracker id coming from the DCF tracker using python bindings?
When I run the deepstream-app I can see the tracking ids (starting from 0 and incrementing by one every time a new object is detected and tracked). How can I access that id?
I see that there is a “tracker_bbox_info” field in “object_meta”, but I can’t get information from there, only a org_bbox_coords that seems to be a pointer and cannot be casted.
Any idea? Thanks!

11

Check this,

https://docs.nvidia.com/metropolis/deepstream/python-api/PYTHON_API/NvDsMeta/NvDsObjectMeta.html

property object_id

Unique ID for tracking the object.