search

FangGet / PackNet-SFM-PyTorch

A Pytorch implementation(unofficial) for paper "PackNet-SfM: 3D Packing for Self-Supervised Monocular Depth Estimation"
GNU General Public License v3.0
21 stars 7 forks source link

how about speed? #1

Closed Mrils closed 4 years ago

Mrils commented 4 years ago

Hello: Thanks for your excellent work! But I am curious how much frame rate the model can reach after reducing the amount of parameters?

FangGet commented 4 years ago

Parameter reduce in this project is not for speed purpose. I will test it later and give the framerate for you.

FangGet commented 4 years ago

@Mrils I just had a test for only depth estimation part on both CPU and GPU, here are the results: for GPU (RTX 2060), it's on average 10ms per input frame (640 192) for CPU (i7-8700 3.2GHz), it;s on average 30s per input frame (640192)

Hope it helps.

Mrils commented 4 years ago

thx!