search

jac99 / MinkLocMultimodal

MinkLoc++: Lidar and Monocular Image Fusion for Place Recognition
MIT License
100 stars 9 forks source link

How to normalize the RobotCar pointcloud data to [-1,1]? #13

Closed LZL-CS closed 2 years ago

LZL-CS commented 2 years ago

Hi, in your paper "As point coordinates in the Oxford RobotCar dataset are normalized to be within [−1,1] range, this gives up to 200 voxels in each spatial direction." May I know what algorithm was used to get the normalized point cloud between -1 and 1?

jac99 commented 2 years ago

Hi, we used Oxford RobotCar point clouds prepared by the authors of PointNetVLAD paper. They also provided Matlab script generating the data. See: https://github.com/mikacuy/pointnetvlad/blob/master/submap_generation/generate_submaps.m

In lines 189-210 they do this point cloud normalization. Basically, they compute the point cloud centroid and shift all points so that point cloud center has (0,0,0) coordinates. And multiply by some scaling factor so x,y coordinates fell into -1..1 range.

LZL-CS commented 2 years ago

ok, I will try it, thank you very much for your reply.