WHU-USI3DV / CoFiI2P

[IEEE RA-L 2024 & ICRA'25] CoFiI2P: Coarse-to-Fine Correspondences-Based Image-to-Point Cloud Registration
https://whu-usi3dv.github.io/CoFiI2P/
92 stars 5 forks source link

Regarding the final result. #2

Closed LoveMYChen closed 5 months ago

LoveMYChen commented 5 months ago

Hello, I have downloaded your training model and Kitti dataset, and conducted model evaluation. I have obtained some numbers, but the problem now is that I am not sure how to use these numbers. My ultimate goal is to adjust the camera pose using my own dataset. Can you give me some guidance?

kang-1-2-3 commented 5 months ago

The eval_all.py produces the result on rotation and translation error. You can load the files and calculate statistics of them. The calc_result.py give the example for specific threshold of these errors. Hopes that it helps you.

LoveMYChen commented 5 months ago

Can we use some algorithm to calculate the correct pose after obtaining errors in rotation and translation? Do you have any recommended algorithms?

martin-liao commented 5 months ago

Can we use some algorithm to calculate the correct pose after obtaining errors in rotation and translation? Do you have any recommended algorithms?

I can't understand what you meant. The registration error is calculated with predicted rotation/translation parameters and ground-truth rotation/translation parameters. The estimated poses are accurate enough in most scenes.

martin-liao commented 5 months ago

Since the problem has been resolved, I will close the issue.