search

raulmur / ORB_SLAM2

Real-Time SLAM for Monocular, Stereo and RGB-D Cameras, with Loop Detection and Relocalization Capabilities
Other
9.17k stars 4.69k forks source link

How to apply ORB_SLAM2 for the KITTI dataset sequences in color (RGB)? #1069

Open larissasantesso opened 1 year ago

larissasantesso commented 1 year ago

I am in doubt in how to use ORB_SLAM2 correctly with the odometry dataset in color (65 GB) located in this link.

Mainly because we have in calibration file P0, P1, P2 and P3. P2 and P3 are the projections matrices for the cameras in color, using as reference the gray left camera. More details of the setup used in KITTI dataset located in this link.

An example of one calib containing these matrices is shown below. We can observe that they have the same focal lengths and centers, but when we compare the last column of P2 (left color camera) and P0 (left gray camera), P2 have other values that are not used in the config file of ORB_SLAM2. We only use the first value that corresponds to Camera.bf (stereo baseline times fx).

So I think I would have to apply some transformations and maybe in the groundtruth file, but I don't know how.

P0: 7.070912000000e+02 0.000000000000e+00 6.018873000000e+02 0.000000000000e+00 0.000000000000e+00 7.070912000000e+02 1.831104000000e+02 0.000000000000e+00 0.000000000000e+00 0.000000000000e+00 1.000000000000e+00 0.000000000000e+00

P1: 7.070912000000e+02 0.000000000000e+00 6.018873000000e+02 -3.798145000000e+02 0.000000000000e+00 7.070912000000e+02 1.831104000000e+02 0.000000000000e+00 0.000000000000e+00 0.000000000000e+00 1.000000000000e+00 0.000000000000e+00

P2: 7.070912000000e+02 0.000000000000e+00 6.018873000000e+02 4.688783000000e+01 0.000000000000e+00 7.070912000000e+02 1.831104000000e+02 1.178601000000e-01 0.000000000000e+00 0.000000000000e+00 1.000000000000e+00 6.203223000000e-03

P3: 7.070912000000e+02 0.000000000000e+00 6.018873000000e+02 -3.334597000000e+02 0.000000000000e+00 7.070912000000e+02 1.831104000000e+02 1.930130000000e+00 0.000000000000e+00 0.000000000000e+00 1.000000000000e+00 3.318498000000e-03

Thanks in advance!!

Echo-gh commented 2 months ago

Hey, have u solved this issue? I have encountered a similar problem. When running ORB-SLAM2 with KITTI-odometry color datasets, the results became pretty bad. I think it may be resulted from the color images calib data.