Sorry, I found that this was caused by my mistake. The camera I am using is Oak-D-Pro, but the test file I started is https://gitee.com/oakchina/camera-start-up-code-samples/blob/main/oak_d_align_sr.py ， They don't match each other.
But, I have another question ：has the Oak-D-Pro camera implemented image time frame synchronization on hardware? When I use this camera, I feel that the depth output effect is not good，especially in desktop scene.
I attempted to adapt RTAB-Map for VSLAM using this camera, employing two architectural approaches:
one being RGBD odometry:
Another type is stereo-inertial odometry:
However, neither effect works well. I have already adjusted many settings and parameters according to the documentation (even switching the baseline reference camera from the left to the right camera), but there is a common issue: when I slowly pan the Oak camera up, down, left, or right while holding it, the odometry output remains relatively normal and meets expectations. However, as soon as I rotate it, regardless of speed, the SLAM immediately starts tracking, jumps, and loses data, then reports an error: "The depth disparity between the left and right cameras is too large to match.".
This also aligns with the phenomenon I observed with my eyes—the depth effect is not good.
OskarSonc
