Open Pedrous opened 6 years ago
I solved the 2nd problem by pre-rectifying the images. The extrinsics are condensed into a single parameter, "baseline", with this implementation. There is no way to pass a general R,t stereo calibration unfortunately (parallel looking cameras are assumed).
Thank you, which parameters do you use then to make sure that the rectification is done only by you? I was able to get very good result with one camera by DSO but now that I am trying to use this stereo implementation, I am facing some difficulty to provide a reasonable trajectory.
I just use the format described in the Readme under "Calibration File for Pre-Rectified Images". But, I only tested it with kitti dataset, no other configurations.
hello,,would u tell me how to set the camera.txt if I use a stereo camera?? I don't know why the width and height are classified by in and out,,,what's the meaning of them? thank u~
It is just if you need to adjust the image size. The image size needs to be such that you can divide it for enough pyramid layers.
So you do not need the extrinsic camera calibration parameters at all, How is the initialization done? And I am wondering if the intrinsic parameters for the second camera are not used or the two cameras are assumed equivalent?