Description
Dear ORORA authors,
First of all, thank you for releasing this impressive work and making the code publicly available. I am new to SLAM algorithms .I have been evaluating the ORORA system using the MulRan KAIST03 sequence, and I’ve encountered a couple of issues I’d appreciate your guidance on.
1. High Estimation Error Compared to Reported Results
While running the evaluation on KAIST03, I obtained the following results:
| Method | ATE (m) | Translation Error (%) | Rotational Error (deg/100m) |
|---|---|---|---|
| ORORA | 573.53 | 118.46 | 15.14 |
| RANSAC + MC + Doppler | 518.79 | 119.36 | 15.09 |
| RANSAC + MC | 526.56 | 120.07 | 15.07 |
These are significantly higher than the results reported in your paper. I followed the standard procedure as per the README and used the provided config files and dataset splits.
2. Ground Truth Timestamps Missing
Additionally, I noticed that the ground_truth.csv file provided for MulRan does not contain timestamps that match the image or radar frame timestamps. This makes it unclear how pose alignment is expected to be done for accurate evaluation.
Any help or clarification would be greatly appreciated. Thank you again for your contribution to the community!
Warm regards,
Lakshmi