I’ve testing the NVDCF tracker which so far seems far better at keeping track of objects than the KLT tracker. However I have noticed a strange situation that I’ve seen happen twice now… I have been walking around in front of the camera and then left the scene… The bounding box then gradually grew and grew over the course of about 30 - 45 seconds until it reached the entire frame size and it stayed there. It would not disappear until I restarted the app.
I am using the tracker config file that is in the
deepstream-nvdsanalytics-test app. How can that happen? What parameters would I look at changing to avoid this issue or a bounding box stuck forever??
• Hardware Platform (Jetson / GPU) Jetson Nano
• DeepStream Version DS 5.0
I’ve noticed sometimes it gets stuck without going to full frame size as well.
If the pgie gets a false detection where it thinks there is a person in the trees the tracker can lock onto it forever until I stop the program and start it again…
Anyone from nvidia able to help. The new ds5.0 dcf tracker can get stuck when a tracker object quickly leaves the scene. It’s like it get stuck in an infinite loop with the bounding box growing to fill half the frame - sometimes the while frame.
It will never die and won’t go away until another real object enters the scene and start getting tracked.
I am using the config file as used by the analytics sample app.
The lingering problem will be resolved if you add the following field in
tracker_config.yml This additional param will be added when we release DS 5.0 GA.
Your issue is very similar to an existing topic, so it would be helpful if you take a look at here; Deepstream 5.0 nvdcf tracker behavior
Hope the existing comments/answers to that thread is also applicable to you. Please let us know if you have any further questions.