search

dongheehand / DeblurGAN-tf

Tensorflow implementation of DeblurGAN(Blind Motion Deblurring Using Conditional Adversarial Networks)
80 stars 24 forks source link

train loss #7

Closed ylhua closed 4 years ago

ylhua commented 5 years ago

why the content loss is so big when training?

dongheehand commented 5 years ago

I used tf.reduce_sum(*) function for content loss ( 154-th line in Deblur_Net.py ) It is because of that function.

I think if you replace tf.reduce_sum() to tf.reduce_mean(), it could be better. (I will modify the code soon!)