ZJULearning / pixel_link

Implementation of our paper 'PixelLink: Detecting Scene Text via Instance Segmentation' in AAAI2018
MIT License
768 stars 255 forks source link

loss = 0.0000 #168

Closed benlee73 closed 3 years ago

benlee73 commented 3 years ago

image

I don't know why loss is zero. The training worked well with the icdar2015 dataset, but when I used my other dataset image and ground truth, I got this result. The format of ground truth is the same as icdar2015. How can I fix the problem?

For reference, my image size is about 601x792. And i'm not change another config or something.