test result error - Githubissues

JiahuiYu / generative_inpainting

DeepFill v1/v2 with Contextual Attention and Gated Convolution, CVPR 2018, and ICCV 2019 Oral

http://jiahuiyu.com/deepfill/

Other

3.27k stars 787 forks source link

test result error #470

Closed wangshuo2020 closed 3 years ago

wangshuo2020 commented 4 years ago

add_res case2_mask case1_output

Why is the training result like this? I tried several times, but no good results. I used the places2 pre-training model.

JiahuiYu commented 3 years ago

I think the mask you provided is difficult, in terms of leaked person legs. I think it would work better for object removal, which is the main use case for inpainting.