Object counting and information extraction

Hello, I’m a beginner using Deepstream, I have DS5.0 installed on my computer alongside with TensorRT 7.0 and I’m using ubuntu 18.4 .My GPU is GTX1050 4GB .
After training and deploying my tlt model I want to extract some information like detected objects (class_name,tracking_id) also I want to count detected objects seperately .
Any help will be appreciated.
Thank you,

Hi,
Take a look at deepstream-test1 application under …/deepstream-5.0/sources/apps/sample-apps
It basically counts the total number of vehicles and people for each frame

Hi @ekimku, do you have idea how to count the vehicles but NOT for each frame? I mean kind of if a vehicle appears on the scene then count and add +1 to a counter but this for all the time when these vechicle is present on the scene. If I count the vehicle on each frame then the count will be fail becasue the number will be for the same vehicles (taking on each frame)…

Regards.

Hi, i coudn’t understand what you mean, do you want to count the total number of unique vehicles in whole stream/video? If yes, you need to use a tracker

Yes, you are right, I would like to count the total number of unique vehicles in whole stream/video.
So I guess I need to read the documentation about Tracker … do you have some recommendation or sample code for check how to use it?

Thanks in anvaced.

Well your task is very difficult since you need to identify each unique car. Tracker basically gives id to each new object in the stream, so the final id that is assigned by tracker should indicates the total number of unique objects.
But the thing is tracker is not very accurate.
You can learn more about tracker from plugin manual
https://docs.nvidia.com/metropolis/deepstream/plugin-manual/index.html#page/DeepStream%20Plugins%20Development%20Guide/deepstream_plugin_details.html#wwpID0E0540HA