I am doing depth experiments to understand the behavior and limitations. To assist, I started showing the output of the left and right cameras. I noticed that horizontal objects were lower in the right camera than the left camera (I measured about 6 pixels for an object about 2 meters away). I have positioned the bottom of the OAK-D level and the left/right camera openings at the same level, to the best of my ability. I worry that my OAK-D is defective (the right inner housing seems to be rotated relative to the left, but that may not mean anything). However, I can speculate that the "misalignment" could be intentional to help calculate depth in the y direction (I admit that is probably not the case). Can anyone offer help? Thanks.

I should have mentioned I was using 640x400 pixels for the mono resolution.

I think I must apologize for acting too hastily. After deeper reading and experimentation, I surmise that the output of the cameras is not used to measure disparity; instead, the rectified left and right images are used. In the latest version of DepthAI it is possible to show the rectified images. There appears to be no difference in the height of horizontal objects in the rectified images. Good!

However, now I think I see another seeming anomaly. In my experiments, I also display the color camera preview and center the color camera on a vertical edge. The mono cameras see the edge at different locations, as expected; it is to the right of the middle in the left image and to the left in the right image. However, in the left rectified mono image the edge is in the middle of the image. And in the right rectified mono image, the edge if farther to the left. And finally the edge that should be in the middle of the depth output is instead located at its location in the right rectified image.

I know that disparity is calculated in the right image. I'm not using depth align. I was hoping the default was CENTER, but that seems not to be the case. So the default must be RIGHT. Does this make sense?

    Hello gregflurry , correct, rectification process hugely eliminates problems you have described in first post. The default is right camera, since that's the least computationally expensive - aligning depth frame to eg. color camera takes quite a lot of processing resources. Hopefully, that answered your question.
    Thanks, Erik