I am trying to build a multi-camera tracker. All the cameras are connected to a single Xavier AGX, running Deepstream 5.0 with JetPack 4.4.1. I am trying to use DCF as a tracker in Deepstream. I have a system where the targets are sparse, but are moving between camera views. I need to have the same tracking ID when an object moves from one camera view to another.
I am trying to use DCF as a tracker with batch-process turned on. I was hoping that DCF already has this feature built-in. Am I missing something?
Thank you,
Ramya.
Current NvDCF tracker and other tracking algorithm in deepstream do not support tracking object in different cameras. Please implement your own tracking algorithm outside of deepstream.
Thank you for letting me know. Can you kindly elaborate on what batch-process means in nvDCF?
‘batch-process’ is for parallel calculation with GPU. Batch means a group of data, for deepstream’s case, it means a group of frames. With Nvidia’s GPU, the video/image procesing can be applied with multiple frames simultaneously. So we call it ‘batch-process’.
It is not specific for NvDCF.
Please search batch and parrellel compute in internet.