Noise-control-multi-object-tracking

A plug-and-play solution mainly focuses on the noise-control designs of detection, association, and reconnection.

Installations

Python3
Pytorch 1.4+ or higher

Module

Heatmap Feature Prior Denoiser

HFPD is applied to the features after the addition of heatmap and image fusion, such as the DLASeg part of dla.py in CenterTrack to highlight the central point features of heatmap.

Smoothing Gain Kalman filter

SG Kalman is applied to the Kalman filtering algorithm of data association, and it takes effect when the confidence is passed into the function project.

Gradient Boosting Reconnection Context

GBRC is used to reconnect the debris track after obtaining the result. Put the result into the tmp folder, run GBRC.main, and obtain the reconnection result in the interresult folder. You need to change the name of the new file in opts.

Citation

@article{you2023NCT,
title={NCT:noise-control multi-object tracking},
author={Zeng, Kai and You, Yujie and Shen, Tao and Wang, Qingwang and Tao, Zhimin and Wang, Zhifeng and Liu, Quanjun},
journal = {Complex & Intelligent Systems},
year = {2023},
doi={https://doi.org/10.1007/s40747-022-00946-9 }
}