Re-Identification across cameras or re-identification in same camera

[Duggy]

Hi,

I am running the demo model for the multi camera multi tracking solution in python. This demo uses reidentification. What I see happening is that in the same camera when the person is identified they are given the ID throughout the view (most of the time). However when they leave the camera and then return to the view they are given a new ID and tracked through the view.

Are they not supposed to receive the ID from before? Is this not re-identification? So if they get ID 10 when they are in the view, leave the camera and then return they should receive ID 10 again? Or am I mistaken?

Thanks

[Zulkifli_Intel]

Hello Elad Segev,

Thank you for reaching out to us.

Yes, you are right. For instance, when camera 1 detects object 1, then it will assign as ID0. When camera 2 detects object 1, it will make a comparison with the assigned ID in camera 1. If matched, it will assign object 1 as ID0. The reidentification is also a difficult task because the appearance features of objects may change dramatically on account of varied illuminations and viewing angles in two different cameras.   

You can refer to person-reidentification documentation and Multi Camera Multi Target Python Demo documentation for more information.

Sincerely,

Zulkifli

[Zulkifli_Intel]

Hello Elad Segev,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.

Sincerely,

Zulkifli