Calibration Performance - İmplementation on KITTI Dataset about joint-lidar-camera-calib HOT 2 CLOSED

Thanks for your interest in my work!

1. Sure. I can send you the demo code for running our algorithm on the KITTI dataset.
2. Indeed, it does not change anything. The camera_coeffs in config/config.yaml only serves as the reference (for instance, intrinsic parameters calibrated by standard chessboard-based methods), to indicate the accuracy of our method. If you are pretty confident in your intrinsic parameters and do not want to optimize them, set keep_intrinsic_fixed to true in config/config.yaml.
3. You can send the data which produce exactly the same result to my email ([email protected]). I will take a look.
4. I will give my answer after seeing your collected data.
5. Several months ago I also encountered this problem on KITTI dataset. However, the solution is straightforward. Simply change two parameters in COLMAP and everything will be fine. I will detail the solution soon.
6. Frankly speaking, I do not know either. While intuitively the more data, the better, I have not conducted experiments to verify this.

from joint-lidar-camera-calib.

Increase init_max_forward_motion to 1.0 and decrease init_min_tri_angle to 5.0.
This should help.

from joint-lidar-camera-calib.