About the efficiency
riyueshi opened this issue · comments
Hi,
The depth map present in the paper is really amazing, I wonder how much time it cost to produce a single depth map?
Thanks,
Miller
A forward pass of the proposed network takes about 0.5 seconds for 2-view matching and an additional 0.25 seconds for every new frame matched (640 × 480 image resolution).
That's great, and what about the hardware of this performance
I test it on the computer (i7-6800K CPU, one GTX 1080Ti).
Thank you!
adding
cudnn.benchmark = True
to test.py improved the performance from ~4s to ~.7s on a 1080Ti