Reference Papers and Parameters explanations about sdr

[TingmanYan]

Yeah, I am glad to, but the paper is still under review.
The review will be finished in a few weeks. Please wait for sometimes if you like.

[cumtchenchang]

@TingmanYan Thank you for your good work, and I am very interested in your paper. When your paper is accepted, please update your work. Good luck for you. Thank you.

[luomomo]

Yeah, I am glad to, but the paper is still under review.
The review will be finished in a few weeks. Please wait for sometimes if you like.

Thanks for your work. I can see in the paper it performs good. So I want to test in my picture.Do you have any experience about using your algorithm with other data set because I see you have special operation with KITTI and Middv3 ?I want to test your refinement method with my own picture, but the graph cuts part always crash in memory ( new and delete)

[TingmanYan]

What's the resolution of your image and the memory of your computer? Since the graph cuts operates on superpixels, the memory shall be enough.
The KITTI and Middv3 modes are mainly for benchmarks. You can also directly invoke the method as:

        FastDR fdr(im0, disp_WTA, params, maxdisp, 0);
        cv::Mat labeling, refined_disp;
        fdr.run(labeling, refined_disp);

where im0 is the color reference image and disp_WTA is the WTA disparity map. Make sure the maximum disparity of disp_WTA is less than maxdisp.

[danielcranston]

@luomomo make sure the disparity maps you're using contain positive disparities and that your calib.txt reflect your images (heigh/width/ndisp).