search

SRainGit / CAE-LO

CAE-LO: LiDAR Odometry Leveraging Fully Unsupervised Convolutional Auto-Encoder for Interest Point Detection and Feature Description
113 stars 28 forks source link

Demo data does not contain ground truth poses #10

Closed nanaBaah closed 4 years ago

nanaBaah commented 4 years ago

I have 2 question:

  1. Can i get a snippet of the folder structure for the project. Maybe a tree format of it?

  2. The DemoData does not contain all the snippet of the data required to run the code. Can I get the data for poses, poses__, poses___ ? and any other sample data you can share.

Thanks in advance

SRainGit commented 4 years ago

Yeah, DemoData doesn't show the whole folder structure, but other folders like 'KeyPts' and 'VoxelModel' etc., are just under the same base folder with 'Features'. A tree structure will be:

sequences
--00
----Features
----Features-3DFeatNet
----Features-USIP
----InlierIdx
----KeyPts
----RefinenmentData
----SphericalRing
----velodyne
----VoxelModel
----calib_.txt (removed texts from the original calib.txt)
--01
----......
--02
......

DemoData is just to visualize the matching comparison. Actually running the code only needs the benchmark data from KITTI. poses is the ground truth, and you can download it from KITTI website; poses_ is the initial odometry generated by frame-by-frame matching; poses__ is the de-jumped poses of poses_ (not shown in the paper, but you can check the code, very simple); poses___ is refined poses based on poses__; To get the refined poses, you need to run through the above steps. The data of poses___, please see the second item of section 'Notes'. As for the data of pose_ and poses__, I didn't backup them at the time I uploaded my result to KITTI. And then they were covered by my later results. But it is not important. You can check the poses_ of 00 and 01 in DemoData.