[Help]Using Unet to color the Japanese animation picture

Part 1 (2019)
1

image

Which lr should I pick?
I still think that I am not doing great. Do any of you have a suggestion that I can read and code it better to improve it?

I don’t know how Jason did his project, but I am using lesson 7 Jupyter notebook to do it.
It looks like it is using imagenet to pre-train it. Should I use my own model that I was using to train the same anime charterer as a pre-train model? I used it in lesson2.
https://github.com/JonathanSum/Pytorch_Projects/color_comic_happy_sugar_life(WIP)_.ipynb

dataset:https://github.com/JonathanSum/Pytorch_Projects/blob/master/Happy%20Sugar%20Life_Deep_Learning/color.zip

image
image1316×982 296 KB

2

You can try reading papers on the topic, there are a few such as this one

3

I hate reading paper. I hope there will be a online tutorial.:crossed_fingers: