How to Fine-Tune XRDSLAM on KITTI Dataset? + Documentation Inquiry & Feature Suggestion

[akramsalim]

Hi team,

Thank you for developing XRDSLAM—it’s an excellent tool! I have a few points to discuss:

1. How can I fine-tune XRDSLAM using the KITTI dataset? I’m interested in adapting this tool for KITTI's outdoor SLAM scenarios. If possible, could you provide guidance or steps for fine-tuning on KITTI?

2-Documentation Availability: I noticed there isn't much documentation available for XRDSLAM. Are there any detailed resources or guides beyond the GitHub README? It would be helpful for understanding the tool’s full capabilities and customization options.

3-Feature Suggestion: If KITTI is not currently supported, would you consider adding official integration for it? Given its widespread use in SLAM and autonomous driving research, it could greatly extend XRDSLAM’s applicability.

Thank you for considering this, and any advice or support would be appreciated!

[wangxiaomeng030]

XRDSLAM is a general algorithm framework that currently supports running SLAM based on existing depth models but does not yet support training specific models for individual algorithms. The KITTI dataset format is not currently supported in XRDSLAM, but by modifying ‘’slam/common/datasets.py‘’, it can be easily adapted to support KITTI in XRDSLAM.

[wangxiaomeng030]

The flowchart in the document “docs/adding_a_new_algorithm.md” provides a detailed overview of XRDSLAM, offering a clear understanding of the entire algorithm framework.