Skip to content

README.md

Latest commit

 

History

History
executable file
·
83 lines (44 loc) · 2.24 KB

README.md

File metadata and controls

executable file
·
83 lines (44 loc) · 2.24 KB

Preparation before training and test

This repo currently offers support for KITTI, KITTI360, Apollo, MulRan, and Ford campus datasets for continuous registration testing; KITTI and KITTI360 for closed-loop registration testing; and KITTI and Ford campus for loop detection testing.

Change output path _C.output_root to your preferred path. Logs, trained snapshots, processed features will be saved in this path.

Dataset (KITTI)

Download the data from the Kitti official website.

--(your dataset path)--Kitti--sequences--00--velodyne--000000.bin
                           |              |         |--...
                           |              |...
                           |--calib--sequences--00--calib.txt
                                             |   |--times.txt
                                             |--...

Change _C.data.dataset_root in experiments/lcrnet/config_model.py to your dataset path

Downsample

Downsample the point cloud:

python data/Kitti/downsample_pcd_kitti.py

Generate ground truth continuous pairs

Generate ground truth text for registration testing:

python data/Kitti/generate_kitti_pairs.py

You can also use our generated ones in assets/data/registration/icp10.

Put the files in (your dataset path)/Kitti/icp10 directory.

Generate ground truth closed loop pairs for registration

Generate ground truth text for loop closing testing:

python data/Kitti/generate_kitti_loop_pairs.py

You can also use our generated ones in assets/data/registration/loop_distance4 and assets/data/registration/loop_overlap0.3.

Put the files in (your dataset path)/Kitti/loop_distance4 and (your dataset path)/Kitti/loop_overlap0.3 directory.

Ground truth loop pairs for loop detection

Please follow overlapnet to generate the ground truth files. Also, you can use our generated one in assets/data/loop_detection/overlap.

Put the files in (your dataset path)/Kitti/overlap, and unzip overlap-based_gt_pairs.zip.

Dataset (KITTI360)

TO BE RELEASED

Dataset (Apollo)

TO BE RELEASED

Dataset (Ford campus)

TO BE RELEASED

Dataset (Mulran)

TO BE RELEASED