search

tsinghua-rll / VoxelNet-tensorflow

A 3D object detection system for autonomous driving.
MIT License
453 stars 123 forks source link

How to get 'lidar_coord'? #16

Closed Zehaos closed 6 years ago

Zehaos commented 6 years ago

Hi, @jeasinema

Could you kindly shed more light on "lidar_coord"? How to get the coordinate of the lidar? Is it necessary to shift the point cloud by adding "lidar_coord"?

jeasinema commented 6 years ago

Hi, I use lidar or camera for coordinate param in some functions located in utils/utils.py just because there are 3 coordinate systems in KITTI dataset, and I use 2 of them. All the lidar points are under the lidar coordinate system, but all the labels are in camera coordinate system. So for this purpose I have to transform the coordinate between them. If you want to get a coordinate under a specific coordinate system, just take a look at here and here.

Zehaos commented 6 years ago

Sorry for my unclear question.

I mean that why we need to shift the point cloud here? To my knowledge, to transform the camera label to lidar coordinate, we only need to use calibration matrix.

Another question, is it suitable to use the mean calibration matrix from training set (related code)?

jeasinema commented 6 years ago

Hi, I think you need to check the 3.1 of the paper. We add the bias only for a nonnegative coordinate, since then the coordinate is used for creating the voxel block. For the second question, I don't know, but I've checked the variance of all the calibration matrixes, it is not so big.

Zehaos commented 6 years ago

Got it, thank you for your patience.

luckysouthchou commented 4 years ago

good job！