From your code, I found the 'interpolation' step for post-processing (mentioned in section 4.3.1 in your paper 'Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches') is only for 'dataset == kitti' or 'dataset == kitti2015'. So if 'dataset == mb', will this step be skipped? If so, why not doing this for middlebury datasets as well?
Thank you very much!
Hope to hear from you soon. @jzbontar
Dear Mr. Žbontar,
From your code, I found the 'interpolation' step for post-processing (mentioned in section 4.3.1 in your paper 'Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches') is only for 'dataset == kitti' or 'dataset == kitti2015'. So if 'dataset == mb', will this step be skipped? If so, why not doing this for middlebury datasets as well?
Thank you very much! Hope to hear from you soon. @jzbontar