search

tinghuiz / SfMLearner

An unsupervised learning framework for depth and ego-motion estimation from monocular videos
MIT License
1.96k stars 555 forks source link

Why pose GT you provided differ from KITTI site? #149

Open murraiya opened 7 months ago

murraiya commented 7 months ago

Hi I am a newbie in this field and trying to run SfMLearner. Your job is incredible. Thank you for sharing code.

I have two question.

  1. I checked out KITTI odom pose GT is differ from your GT provided in eval section. Can you explain why? Is this involved with coordinate system or sth?
  2. I evaluated my trained model with your GT and got ATE: 0.013, std:0.0098 which is better than in paper. Did I do something wrong? + I am working out but I can't understand how depth involves in pose net.

Thanks.