Reconstruction Stuck at Original Position in MonoGS

[tensorpura]

Hi, thanks for the great work!

I encountered an issue where the reconstruction either stays in the original scene or significantly lags behind the actual camera position, even after multiple frames of input (over 400, Indoor scenes). The GUI is continuously updating, and new pose information is being input, but it always lags behind. This causes the Gaussian ellipsoids to appear chaotic and misplaced in the scene.

Context:
I processed my custom dataset following the Replica dataset format.

The input includes depth maps, RGB images, pose information, and camera calibration parameters, all aligned one-to-one.

Problem:
Despite providing these inputs, the system fails to track the camera properly, either not updating the reconstruction or lagging behind as new frames are processed. I suspect this could be related to image matching or tracking issues.

Questions:
Are there any specific requirements or constraints for the input dataset that I may have overlooked?

Is there additional documentation or guidance on data preprocessing and configuration that could help resolve this issue?
Any advice or suggestions would be greatly appreciated. I am happy to provide further details or logs if needed.

Thank you for your time and support!