MTGS uses datasets in the form of colmap

Thank you again for your excellent work. I encountered some problems when running my own dataset using the MTGS code. My reconstruction failed when I only had image, depth, sfmpoint, cam2world, and other information. The trajectory information looked three-dimensional, but I only had x, y. And I don’t have ego information. How should I use MTGS in this case?