What does the loss function's value finally converge to? - Githubissues

gouthamvgk / SuperGlue_training

This repository contains the training code for SuperGlue. Uses COCO dataset by generating random homographies.

Other

101 stars 22 forks source link

What does the loss function's value finally converge to? #15

Open zshyang opened 1 year ago

zshyang commented 1 year ago

Hi,

I am training on my dataset. And I am using the same loss function. I saw the loss begin at 3.xxx and gradually drop to 1.9xxx. What is your experience with the scope of loss values?

Best regards