search

noahzn / Lite-Mono

[CVPR2023] Lite-Mono: A Lightweight CNN and Transformer Architecture for Self-Supervised Monocular Depth Estimation
MIT License
527 stars 58 forks source link

About the loss #103

Closed wangjiyuan9 closed 8 months ago

wangjiyuan9 commented 9 months ago

Hi, Author！Thank you for your amazing work! I wonder the training loss would be at which scale when converged? If you have some training output log, it would be better!!

wangjiyuan9 commented 9 months ago

I read the previous issue, and know that you didn't keep the loss log. So a range is ok. Be like [0.5(epoch 1) to 0.1 (epoch 20)] is ok!

noahzn commented 9 months ago

Hi, I can't remember the numbers. Please note that the depth losses are not accurate. See here.

noahzn commented 8 months ago

I am now closing this issue due to a lack of response.