Revision history [back]

The Ground Truth (GT) for stereo images is made using different tecnology and combining them such as SLAM, IMU etc.. It is created by the dataset authors and is a reference, something that you want to achieve. Is totally fine have worse disparity map, you can attemp to get better result applying post processing (filtering). You can find A LOT of papers regarding disparity computation and filtering.

Regarding how they have obtainedyou can find that in the references, like for Middlebury dataset, you can find further information on their dataset wepbage, while for KITTI dataset you can find firther information here and here

The Ground Truth (GT) for stereo images is made using different tecnology and combining them such as SLAM, IMU etc.. It is created by the dataset authors and is a reference, something that you want to achieve. Is totally fine have worse disparity map, you can attemp to get better result applying post processing (filtering). You can find A LOT of papers regarding disparity computation and filtering.

Regarding how they have obtainedyou can find that in the references, like for Middlebury dataset, you can find further information on their dataset wepbage, while for KITTI dataset you can find firther further information here and here