naoto0804 / pytorch-inpainting-with-partial-conv

Unofficial pytorch implementation of 'Image Inpainting for Irregular Holes Using Partial Convolutions' [Liu+, ECCV2018]

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Blurry results

wuhaozhe opened this issue · comments

Hi, thanks for your project.
I've trained your model on a face dataset.
After 40K iterations, the result is still blurry.
image

Here is the loss log
image
image

I'm wondering is that normal? How long will it take to get good results?

I'm not sure how the progress goes in your faces dataset (in Places2, 160k is at least OK).
Since this implementation is based on the paper published four years ago, you might try recent methods if you want to get better results faster.

Hi, I want to train my dataset use this model, But this code is about the Pales2 dataset, so how do I train my own dataset?

You can get your own dataset and modify places2.py to load the images in it.