Need some advice for training keypoint detector with camera pose supervision

[Merical]

Hi! Great work on using ego-motion to train the local feature and the depth jointly. I'm trying to train a network under the supervision of camera poses in MegaDepth. But it seems to be hard to converge. Would you please provide some advice? Maybe on WeChat? My personal email is [email protected].

Thank you!

[RaresAmbrus]

Hi, sorry about the slow reply - we didn't try training with pose supervision as we train from videos in a self-supervised way. But it sounds like an interesting idea, what kind of performance issues are you running into?