Multi-camera Tracking with DeepStream DeepSORT features

I just have a looked at this demo which used visual features for multi-camera object tracking:

https://demos.ngc.nvidia.com/metropolis_multicam?ncid=so-link-639233#cid=em21_so-link_en-us

I was wonder is there any way to extract the features from the DeepSORT tracker to use them for multi-camera ID assignment?

What do you mean by multi-camera ID assignment?

I want to use DeepSORT visual features to maintain same ID for same person in a multi-camera setup. for example if a person has ID #1 in one camera if the person go to the second camera view it should still gets the same ID.
Is there any way to extract the ReID network features from the DeepSORT tracker?

1 Like

Output ReID of DeepSORT tracker is in DeepStream roadmap. Thanks.