How nvtracker to deal with batch data？

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

hi
I set nvtracker in test3，the input is more than one stream but How nvtracker to deal with batch data？
deal with one by one or Parallel Operations？
in cpu and gpu the same？
Can you tell me the details，thanks a lot

2

The nvtracker low level library uses a CPU based implementation of the Kanade Lucas Tomasi (KLT) tracker algorithm.
It receives batched input GstBuffers with nvstreammeta and handles them in batched.

You can refer to Deepstream 3.0 Plugin Manual, 2.2 → GST-NVTRACKER

3

hi ChrisDing,thank a lot!
but the details I still don’t quite understand~
1.handles them in batched mean that it is a four dimensional blob，like yolo Plugin(dsPreProcessBatchInput make vectorcv::Mat*& cvmats to cv::Mat& batchBlob)，or a loop to deal one by one?
2.I read Deepstream 3.0 Plugin Manual, 2.2 → GST-NVTRACKER it can configure GPU device gpu
gpu is only for internal scaling/color format conversions,the main Kanade Lucas Tomasi (KLT) algorithm is in cpu is that right?

4

Hi！
is nvtracker with IOU same as KLT handles them in batched？
handles in batch means four dimensional blob？

5

the nvtracker will help to track the objects in each stream separately, run on CPU, you can get the separate result after nvstreamdemux plugin.