Thank you for sharing your work. I have been running multiple experiments using your implementation.
Qualitatively, the visualization results appear to improve after optimization. However, during quantitative evaluation, I consistently observe a systematic issue: the estimated trajectory exhibits a noticeable tilt, and the point cloud shows an upward drift along the Z-axis starting from the first frame.
This behavior is reproducible across almost all of my test sequences.
I would like to ask:
Have you encountered similar trajectory tilt or Z-axis drift issues in your experiments?
If not, could this be caused by improper parameter settings or configuration?
Are there specific parameters (e.g., related to optimization, weighting, or initialization) that are particularly sensitive and should be carefully tuned?
Any insights or suggestions would be greatly appreciated. Thanks again for your work!
Thank you for sharing your work. I have been running multiple experiments using your implementation.
Qualitatively, the visualization results appear to improve after optimization. However, during quantitative evaluation, I consistently observe a systematic issue: the estimated trajectory exhibits a noticeable tilt, and the point cloud shows an upward drift along the Z-axis starting from the first frame.
This behavior is reproducible across almost all of my test sequences.
I would like to ask:
Have you encountered similar trajectory tilt or Z-axis drift issues in your experiments?
If not, could this be caused by improper parameter settings or configuration?
Are there specific parameters (e.g., related to optimization, weighting, or initialization) that are particularly sensitive and should be carefully tuned?
Any insights or suggestions would be greatly appreciated. Thanks again for your work!