An end-to-end tracking framework via multi-view and temporal feature aggregation [paper]

@article{YANG2024104203,
title = {An end-to-end tracking framework via multi-view and temporal feature aggregation},
journal = {Computer Vision and Image Understanding},
volume = {249},
pages = {104203},
year = {2024},
issn = {1077-3142},
doi = {https://doi.org/10.1016/j.cviu.2024.104203},
author = {Yihan Yang and Ming Xu and Jason F. Ralph and Yuchen Ling and Xiaonan Pan},
}

More information：Notes of MultiviewX_Perception

Args

MultiviewX_Perception可以接受命令行参数，从而用户可以快速高效地产生数据集。当未接收到相关参数时候，不会启用相关功能。

-a ：Annotate and show the bbox on the first frame of each camera or not.

-s ：Save the bbox on the first frame of each camera or not.

-k ：Keep the remains of Perception dataset or not.

-f ：Force calibrate and generate POM, regardless of perception.

-p n: Provide preview for the front n frames. ex. -p 5 will provide 5 frames to preview

-v ：Generate Overlap view for the dataset

-view path ：Generate Overlap view for the specified dataset, there should be folder calibrations in the given path. ex. -view D:\Wildtrack