Running my own data

[IsakIngebrigtsen]

Hey, Great work done on this model. I plan to use it as part of my master's thesis, To test whether it can be used to improve navigation in a vehicle by doing point cloud registration between a georeferenced source point cloud, and a standalone target point cloud. In this scenario, I need to use my own data, and also global/absolute coordinates. Do you have any tips to implement XYZ data in the same format as kitti or nuscenes and use it on the already trained model, also are there any constraints in the model when it comes to using global coordinates?

Kind regards
Isak Foss Ingebrigtsen

[FanLu97]

Hello, thanks for your interests.
By using global coordinates, do you mean that the coordinate system is not under the vehicle coordinate system? If true, I recommend that you crop out the part of the point cloud that roughly overlaps with the current one that needs to be aligned and then recenter it by minus the center coordinates. The resulting point cloud should be more similar with the training data.

[nuo-1]

Hey, Great work done on this model. I plan to use it as part of my master's thesis, To test whether it can be used to improve navigation in a vehicle by doing point cloud registration between a georeferenced source point cloud, and a standalone target point cloud. In this scenario, I need to use my own data, and also global/absolute coordinates. Do you have any tips to implement XYZ data in the same format as kitti or nuscenes and use it on the already trained model, also are there any constraints in the model when it comes to using global coordinates?

Kind regards Isak Foss Ingebrigtsen

Hello, have you trained with your own dataset? How about the effect？