ZJULearning / pixel_link

Implementation of our paper 'PixelLink: Detecting Scene Text via Instance Segmentation' in AAAI2018

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Loss around 0.4 ~0.5, and predict empty box

jisheng047 opened this issue · comments

After training about 20k iters, my loss still around 0.4 ~ 0.5. And pick up the model to predict, what i receive is empty box?

Can anyone know how to fine tune the model? Please help me out this problem, thanks you

@jisheng047 how did you solve

@ajinkya933 Keep training until 250k - 300k steps, loss will drop out to 0.2-0.3.