Open Yzmblog opened 4 years ago
Hi, I was confused about why in the finetune.py the mask was set to mask = (disp_true > 0); but in main.py it was set to mask = (disp_true < 192) ? What cause this difference? and will it affect the result significantly?
in the paper and devkit tool of http://www.cvlibs.net/datasets/kitti/eval_scene_flow.php?benchmark=stereo ,you can know in the scene flow dataset some pixel' disp is very large, so set disp_true<192 and don't need to calculate the loss;but in the finetune.py, you can see the devkit tool of matlab set disp_true>0 ,and i think there are no large disp pixels.
Hi, I was confused about why in the finetune.py the mask was set to mask = (disp_true > 0); but in main.py it was set to mask = (disp_true < 192) ? What cause this difference? and will it affect the result significantly?