deepstream-bytetracker-terminated-tracks

This project integrates a state-of-the-art ByteTracker into the NVIDIA DeepStream pipeline, modifying both C++ and Python layers to expose terminated track lists.

Byte-Track Integration with Deepstream 6.4

MODIFIED: Integration and instructions updated for Deepstream 7.1 & 6.4, multi-stream and terminated tracks support.

This code is based on the official code.

Integrating Byte-Track C++ code with the Deepstream-6.4.

Features:

• Only support for DeepStream 6.4. For other verions, please refer to the official code.
• Support multi-streams
• The memory leak issue should be resolved.

Build Instructions

$mkdir build && cd build

$cmake ..

$make ByteTracker

This will create ./lib/libByteTracker.so file which can be passed as the custom low level tracker library to deepstream. To do so just add it to the folder

/opt/nvidia/deepstream/deepstream/lib/

In your deepstream_app_config.txt add the tracker.

[tracker]
enable=1
tracker-width=640
tracker-height=384
gpu-id=0
ll-lib-file=//opt/nvidia/deepstream/deepstream/lib/libByteTracker.so
enable-batch-process=1

Modified/Enhanced Features

Multi-stream support: Each stream gets its own ByteTracker instance. Terminated tracks output: See comments in code and the Python probe file. DeepStream 7.1 support: Update config, see /isee/deepstream/configs/config_tracker.yml

Trying out Multiple Detectors

Feel free to try multiple detectors. You can do this by using https://github.com/marcoslucianops/DeepStream-Yolo

Now adding the tracker lines in the config file you should get your tracker working.

References

1. How to Implement a Custom Low-Level Tracker Library in Deepstream
2. Byte-Track