search

KumapowerLIU / Rethinking-Inpainting-MEDFE

Rethinking Image Inpainting via a Mutual Encoder Decoder with Feature Equalizations. ECCV 2020 Oral
Other
371 stars 50 forks source link

weird results #15

Open crazygirl1992 opened 3 years ago

crazygirl1992 commented 3 years ago

hello，

i use my data to train the network, and the test results are weird,like 图片

it's fill the mask region with the weird pixels ,like noise.why this condition happened? thank you!

Young-Zzz commented 3 years ago

Me too... I don't know how to solve this problem.

543877815 commented 3 years ago

1383 My result using pretrain model of celebA....

zzktb commented 3 years ago

@543877815 Something similar to what I've got. I double checked I'm using the CelebA models. 0

LC2978 commented 3 years ago

May I ask, I have trained with my own dataset, and the test results are similar to yours. Have you found the reason

zzktb commented 3 years ago

@LC2978 I am using the pre-trained model they provided. It seems that the network is not working normally. I have no idea what would be the reason.

KleinXin commented 3 years ago

Same weird problem

image

htyjers commented 2 years ago

@crazygirl1992 Have you solved the problem？could you help me