Track a person walking or a moving object on freespace obtained by using jetson inference segmentation


In my project, I have to track a person walking or a moving object on a free space.

Using jetson inference segmentation, i was able to successfully identify free space.
Now for my robotic application my task is to track a moving person on the free space.

Note: I have a limitation to use jetson inference only as it was used for object detection, free space segmentation.

Now i have to extend the task to track person or a moving object on free space detected above.

Is there any way, or i should say is it possible to achieve above?

Language used: Python
Platform: Jetson Xavier

Or is there any way that i can collaborate with any other available tracking algorithm with jetson inference?

Any guidance or github link to explore would be really helpful to achieve my goal.

Thanks and Regards,
Udaykiran Patnaik.


A straightforward way is to train an object detector with jetson-inference.
And check the bbox is within the freespace or not directly.


Hello @ AastaLLL,

Thank you so much for your reply.

Could you please correct me if my below understanding is correct or not?
Step 1:
I shall use Jetson Inference detect-net by loading “ssd-mobilenet-v2” model to get the bounding box of all persons in a frame and get their bbox.
Step 2:
I shall find the centre of the bbox of each person detected in every frame.
Step 3:
Use Jetson Inference segmentation for free space detection and get the pixel region for floor(example).
Step 4:
Check if the output of step 2 is within the free space pixel region (output of step 3) in that frame or not.

Thanks and Regards,
Udaykiran Patnaik.


Yes. It sounds good.
Please also noted that you can apply transfer learning to finetune the model for your use case.


1 Like