hezhangsprinter / DCPDN

Densely Connected Pyramid Dehazing Network (CVPR'2018)
405 stars 112 forks source link

test results on Epoch_8 are not ideal #41

Closed wantbe2020 closed 5 years ago

wantbe2020 commented 5 years ago

Hello, my dehazing results obtained with the netG_epoch_8 test are not ideal, and even obvious problems have arisen. Then I analyzed the atmospheric light value A according to the output, mainly because the predicted A is generally too small, only 0.05. So, the final output graph is too bright, what is the cause of this?

hezhangsprinter commented 5 years ago

if you are not reproducing the images shown in the readme, please reproduce first.

hezhangsprinter commented 5 years ago

22

This may help.