Releases: ctu-vras/depth_correction
Releases · ctu-vras/depth_correction
RA-L submission
Extended experiments on KITTI-360 (SLAM evaluation)
Ouster bias estimation with point-to-plane distance
v0.0.2
Removed dependency on data package: https://github.com/tpet/data.
Fixed ICP point-to-plane loss.
Full training pipeline on FEE Corridor data set (including poses initialization from SLAM).
Experiments and results code for the paper.
v0.0.1
Depth correction models:
- Linear,
- Polynomial,
- ScaledPolynomial,
- InvCos,
- ScaledInvCos.
Supported datasets (wrappers are implemented in data package):
- ASL laser,
- KITTI Odometry,
- KITTI-360,
- NewerCollege,
- Depth Correction.
Map consistency losses:
-
$\lambda_{1}$ - minimum eigenvalue loss, -
$trace(Q)$ - trace loss, - ICP loss (point-to-point or point-to-plane distances between neighboring points).