NaN prediction
tonmoy-saikia opened this issue · comments
Hi, if I input a depth map containing some 0 values (invalid pixels) I get all NaNs in the prediction. Is this behavior expected? How do you handle invalid depth?
princeton-vl / RAFT-3D
tonmoy-saikia opened this issue · comments
Hi, if I input a depth map containing some 0 values (invalid pixels) I get all NaNs in the prediction. Is this behavior expected? How do you handle invalid depth?