TC_tracker

Tracklet clustering for 2D tracking.

1. Prepare the detection file

Please prepare the detection file with the format that follows MOT challenge https://motchallenge.net/instructions/.

2. Set the directory

Set the directory in demo.m.
Input:
img_path: the directory of the image folder that contains the video sequence.
det_path: the directory of the detection file from the previous step.
seq_name: the name of the video sequence.
ROI_path: the directory of the ROI mask. If no ROI provided, use empty matrix instead.
img_save_path: the directory of the output image after tracking.
result_save_path: the directory of the tracking result. The result follows the UA-Detrac format. https://detrac-db.rit.albany.edu/instructions.

3. Set the parameters

Set the parameters in demo.m.
det_score_thresh: detection score threshold between 0 and 1.
IOU_thresh: IOU threshold for detection asscociation across frames between 0 and 1.
color_thresh: color threshold for detection asscociation across frames between 0 and 1.
lambda_time: time interval cost.
lambda_split: tracklet separation cost.
lambda_reg: smoothness cost.
lambda_color: color change cost.
lambda_grad: velocity change cost.

Citation

Use this bibtex to cite this repository:

@inproceedings{tang2018single,
  title={Single-camera and inter-camera vehicle tracking and 3D speed estimation based on fusion of visual and semantic features},
  author={Tang, Zheng and Wang, Gaoang and Xiao, Hao and Zheng, Aotian and Hwang, Jenq-Neng},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
  pages={108--115},
  year={2018}
}