Understanding accuracy of disparity maps based on ground truth

edit

I am using stereo dataset of KITTI. After generating disparity maps, I hand coded Q matrix (using calibration data that comes in calibration files along with kitti dataset) and then reprojectImageto3D. However, after surfing and researching a lot , I think to understand accuracy of disparity map based on groundtruth data, my next step is to generate 3D points using surface matching or PCL but still confuse.. Attaching depth map image and following is the Q matrix that I have used Matrix Q is= [1, 0, 0, -696.0217; 0, 1, 0, -241.806; 0, 0, 0, 958.3579999999999; 0, 0, 0.01851, 0] I need some insight as I am going in right direction or not? Thanks.(/upfiles/14805413884322232.png)(/upfiles/14805413691477726.png)