[time]
2022-08-03T17:41:51+02:00

[track]
2

[team_name]
CamLoc

[team_institution]
1. Institute of Computing Technology, Chinese Academy of Sciences. 2. Beijing University of Posts and Telecommunications

[logolink]


[team_members]
Yushi Chen, Yue Zhuge, Long Wang, Runze Chen, Haiyong Luo, Fang Zhao

[reference_person]
Yushi Chen

[reference_email]
chenyushi@bupt.edu.cn

[description_short]
Generally, we will consider retraining the mainstream absolute pose regression algorithms to achieve the task of predicting the absolute position, floor, and orientation for every single image directly. If the data is scarce, we will consider using the SOTA image retrieval algorithms. If the data is sufficient and covers the majority of areas, we will use the method based on structure. If the data was photographed in a sequential way and the timestamps are very close between adjacent frames, we consider using the Visual Odometry model based on deep learning.


[description_long_link]
https://1drv.ms/w/s!AnBzqrilHPoE1i7lt8o__eOqll3H

[publish_check_]
true

[results_check_]
true

[data_check_]
true

[pdf_check_]
true

[video_check_]
true