Closed lucky9-cyou closed 9 months ago
When using calibration files from http://filebrowser.ram-lab.com/share/I67uNo6q/20220626_calib/calib/. I got following errors when load yaml:
Another issue concerns projecting the point cloud onto the camera plane. Why use the relative transform matrix between the camera and body IMU as the rotation matrix?
In write_data_to_kitti360.ipynb:
But when i using same transform matrix with following code, I got the wrong results:
The point cloud projections onto the camera were correct for the first few frames, but most of the subsequent frames were basically incorrect: The above is 501.png in 20220219_MCR_normal_00
I believe the issue lies with the calibration file. In write_data_to_kitti360.ipynb, you reference 20230618_calib, but on https://fusionportable.github.io/dataset/fusionportable/, only data from 2022 is available.
The calibration_files has been made public. It seems that the format is made with minor modifications compared with the original calibration files: I deleted the original ```!! opencv-matrix''' Could you please try the new one first?
The calibration_files has been made public. It seems that the format is made with minor modifications compared with the original calibration files: I deleted the original ```!! opencv-matrix''' Could you please try the new one first?
For 2023 calibration files, it is ok. But it is not ok for 2022 calibration files. You need to change the event camera calibration from left to right:
Could you please tell me the difference between the ouster00
and ouster00_undistort
? And why use the relative transform matrix between the camera
and body IMU
to project the ouster00_undistort
point cloud onto the camera plane rather than ouster00
.
And could you please provide fusionportable v2 dataset download link?
LiDAR's points will have distortion due to the moving vehicle. Therefore, I undistort the point cloud using the fastlio2 algorithm, and the point cloud is transformed into the body_imu frame. For the ouster point clouds, they are defined in the ouster00 coordinate system. FusionPortable v2 will be released in two months.
LiDAR's points will have distortion due to the moving vehicle. Therefore, I undistort the point cloud using the fastlio2 algorithm, and the point cloud is transformed into the body_imu frame. For the ouster point clouds, they are defined in the ouster00 coordinate system. FusionPortable v2 will be released in two months.
Got it. Thanks for your reply very much.
maybe you should make that repo public