Offline process of disparity/depth

fraymackers

Hello,

I'm currently recording left, right and rgb encoded video to do offline disparity/depth computation without the camera. I tried traditional disparity computation (with open-cv) but I haven't a good result.

A noticed the disparity/depth map computed on the camera gives better results but I want to do computation offline, based on the left, and right mp4 video. I don't want to use the replayer with the camera connected to my computer because I have to process heavy video files and it will be more efficient on my computer.

Is there a solution to have the computation of the same disparity/depth algorithm but on the computer, on Python?

Thank you in advance

François

erik

fraymackers did you rectify the left/right streams before feeding them into opencv stereo? Please refer to: https://github.com/luxonis/depthai-experiments/tree/master/gen2-stereo-on-host

fraymackers

Hello @erik and thanks for your reply,

Yes I did a rectification but done with cv2.stereoRectify(). I see in the repo you send me, the rectification is done with cv2.warpPerspective(). Is it enough to do a good rectification?

I'm using the Oak-D Pro W.

Thank you

Best Regards;

erik

fraymackers maybe you could share rectified/disparity frames?