search

wh200720041 / iscloam

Intensity Scan Context based full SLAM implementation for autonomous driving. ICRA 2020
Other
559 stars 131 forks source link

FLOAM vs ISCLOAM #3

Closed cgbcgb closed 4 years ago

cgbcgb commented 4 years ago

Hey，I have a question，why your FLOAM（lidar odometry only） is better than your ISCLOAM（with loop closure）in KITTI test dataset（sequence 11 ～ 21）？

wh200720041 commented 4 years ago

Actually, the result of isc-loam is better than floam if you compare the trajectory with ground truth. However, the way KITTI dataset estimates trajectory is a little bit different. They collect short term odometer instead of long term, so that loop closure does not contribute to the result. You may check their website for the source code of odometry evaluation. In the meanwhile, there is some parameter change in the floam.

cgbcgb commented 4 years ago

Actually, the result of isc-loam is better than floam if you compare the trajectory with ground truth. However, the way KITTI dataset estimates trajectory is a little bit different. They collect short term odometer instead of long term, so that loop closure does not contribute to the result. You may check their website for the source code of odometry evaluation. In the meanwhile, there is some parameter change in the floam.

You means that the evaluation method of KITTI can't quantitatively evaluate the real translation or rotation error correctly ?

wh200720041 commented 4 years ago

Note that on KITTI website From all test sequences, our evaluation computes translational and rotational errors for all possible subsequences of length (100,...,800) meters.

The odometry is estimated locally

cgbcgb commented 4 years ago

Note that on KITTI website From all test sequences, our evaluation computes translational and rotational errors for all possible subsequences of length (100,...,800) meters.

The odometry is estimated locally

OK，I got it, Thanks ^_^