search

JUGGHM / PENet_ICRA2021

ICRA 2021 "Towards Precise and Efficient Image Guided Depth Completion"
MIT License
326 stars 46 forks source link

About training #64

Open DongyangHuLi opened 1 year ago

DongyangHuLi commented 1 year ago

Can I train as a whole at once instead of in stages？

JUGGHM commented 1 year ago

I am not sure whether it will work well. The scaling factor upon different loss items should be properly adjusted.