Question about executing faster-lio with LIO-SAM dataset about faster-lio HOT 6 CLOSED

@xiaotaw Did you or Bai Chunge do those LIO-SAM experiments? I'm not quite sure about the details here.

from faster-lio.

1.1 In the paper, liosam_1 is short for campus_small_dataset.bag.
1.2 Yes, the groundtruth of LIO-SAM dataset comes from topic /gps/fix. As rotation is not provided, we just fill them with zeros(i.e. the qx qy qz qw columns in tum file). And the groundtruth trajectory is converted into 100Hz via b-spline interpolation.

2 Due to the limited time, we did not tuned faster-lio for other LIO-SAM's datasets. According to your video, it seems something goes wrong with the bias of the accelerometer, not sure yet.

from faster-lio.

@xiaotaw
Thanks for your reply. 😄

Suppose the result is not good even if I change the parameters associated with the bias of the accelerometer.
Could I understand that faster-lio is not an appropriate algorithm for the rotation dataset?

from faster-lio.

It could be more precise and rigorous to make such a conclusion, if you test on more datasets, and make sure that those datasets are properly fed into the algorithm without any other issues, such as timestamp, extrinsic calibration, etc.

from faster-lio.

@xiaotaw
Thanks for your advice. I'll run tests on the other datasets and look at other issues. 👍

from faster-lio.

1.1 In the paper, liosam_1 is short for campus_small_dataset.bag. 1.2 Yes, the groundtruth of LIO-SAM dataset comes from topic /gps/fix. As rotation is not provided, we just fill them with zeros(i.e. the qx qy qz qw columns in tum file). And the groundtruth trajectory is converted into 100Hz via b-spline interpolation.

2 Due to the limited time, we did not tuned faster-lio for other LIO-SAM's datasets. According to your video, it seems something goes wrong with the bias of the accelerometer, not sure yet.

Hello,
A tiny problem in the paper.
The distance of campus_small_dataset.bag is only ~0.55km, rather than 1.43km which is reported in the paper.
1.43km is for campus_large_dataset.bag.

Best,
Xingyu

from faster-lio.