Closed FlyingGaifield closed 3 years ago
The EPE and time are directly taken from the original StereoNet paper, both are measured by the 8x, multi
model (you can refer to Table 1 and 2 in StereoNet paper).
The EPE and time are directly taken from the original StereoNet paper, both are measured by the
8x, multi
model (you can refer to Table 1 and 2 in StereoNet paper).
From the implementation of https://github.com/meteorshowers/StereoNet-ActiveStereoNet
While using 8x, single,
he can only get 30FPS with epe 1.38. I doubt that 0.015 is not corresponding to 8x, multi
I noticed that in table3, you mentioned EPE/Time of StereoNet is 1.1 and 0.015. I also read the original paper of StereoNet, maybe "0.015s" using the fast(16x, single) version and "1.1 "using the best(8x,multi) version. Can you explain it more?