search

ZM-Zhou / SDFA-Net_pytorch

Apache License 2.0
17 stars 5 forks source link

about Disparity range #2

Closed ksh11023 closed 1 year ago

ksh11023 commented 1 year ago

Hello,

Is there specific reason for setting the disparity range [2, 300] ? Can SDFA-net be trained with disparity range [0,192] ?

Thank you!

ZM-Zhou commented 1 year ago

Hi ksh11023,

We follow the existing work (FAL-Net) to set the disparity range [2, 300]. Of course, you could try to set different disparity ranges in your training (by modifying the yaml file). Please note that the min_disp should not be set as 0, as it would cause a numerical error during depth computing.

ksh11023 commented 1 year ago

Thank you for the reply !