re-train model about paintschainer

[taizan]

Okay thanks for using
I uploaded liner_f to downloads but you can train models without liner_f . ( liner_f is src image to line drawing model. )
For re-training, you need
line image folder
color image folder
image name list file
image name in 2 folder should be same.
and set it in line 67 in train_128.py
dataset = Image2ImageDataset("dat/images_color_train.dat",root+"line/",root+"color/", train=True)

[liamjones]

I'd also be interested to know how big the training data set was for the model you're using

[masato-io]

@liamjones
according to this author's article, he mentioned that this model was trained by around six hundred thousand images.
http://qiita.com/taizan/items/cf77fd37ec3a0bef5d9d

[masato-io]

@taizan
Thank you so much for your quick response. I will definitely experiment re-training with your suggestion. Really appreciated :)

[masato-io]

As my question is clearly & kindly answered by author, I'm closing this ticket.