search

BIT-MJY / SeqOT

[TIE 2022] SeqOT: A Spatial-Temporal Transformer Network for Place Recognition Using Sequential LiDAR Data.
MIT License
101 stars 8 forks source link

convert csv poses to npy poses #3

Open sijieaaa opened 1 year ago

sijieaaa commented 1 year ago

Hi Junyi,

Could you help to provide the code that converts from .csv poses to your processed .npy poses? :)

Thanks!

Regards, Sijie

BIT-MJY commented 1 year ago

Thanks for your interest in our work! I am busy writing my master thesis but will release the code you need in one month.

Your patience is appreciated :-)

ydragon719 commented 1 year ago

Hello! Do you have the conversion code to the npy ? Like the T_w_r_gt_0108.npy.

Yan-Yx commented 1 year ago

Hi Junyi,

I would also like to know how to convert the .csv file of the ground truth of NCLT to the .npy file. Could you please help to provide the code?

Thanks and best regards!

cpkingw commented 1 year ago

It seems that the ground truth (.npy) is obtained by linear interpolation based on the .csv files and the timestamps of .bin. Refer to the read_ground_truth.py file in the dataset website for the implementation of interpolation. As for the 'nan' in the groundtruth.csv of 2012-01-08, they are replaced with the first normal values. The rotation matrix is converted from RPY. However, there is a transformation difference between the groundtruth calculated and that provided (.npy), which confuses me.

BIT-MJY commented 1 year ago

Sorry for the late reply! You could refer to pole-localization to generate the poses at first. The release in this repo is coming.