epochs: 30

output_dir: output_dir

model:

name: AnimeGANV2Model

generator:

name: AnimeGenerator

discriminator:

name: AnimeDiscriminator

gan_criterion:

name: GANLoss

gan_mode: lsgan

# use your trained path

pretrain_ckpt: output_dir/AnimeGANV2PreTrainModel-2020-11-29-17-02/epoch_2_checkpoint.pdparams

g_adv_weight: 300.

d_adv_weight: 300.

con_weight: 1.5

sty_weight: 2.5

color_weight: 10.

tv_weight: 1.

dataset:

train:

name: AnimeGANV2Dataset

num_workers: 4

batch_size: 4

dataroot: data/animedataset

style: Hayao

transform_real:

- name: Transpose

- name: Normalize

mean: [127.5, 127.5, 127.5]

std: [127.5, 127.5, 127.5]

transform_anime:

- name: Add

value: [-4.4346957, -8.665916, 13.100612]

- name: Transpose

- name: Normalize

mean: [127.5, 127.5, 127.5]

std: [127.5, 127.5, 127.5]

transform_gray:

- name: Grayscale

num_output_channels: 3

- name: Transpose

- name: Normalize

mean: [127.5, 127.5, 127.5]

std: [127.5, 127.5, 127.5]

test:

name: SingleDataset

dataroot: data/animedataset/test/HR_photo

preprocess:

- name: LoadImageFromFile

key: A

- name: Transforms

input_keys: [A]

pipeline:

- name: ResizeToScale

size: [256, 256]

scale: 32

interpolation: bilinear

- name: Transpose

- name: Normalize

mean: [127.5, 127.5, 127.5]

std: [127.5, 127.5, 127.5]

keys: [image, image]

lr_scheduler:

name: LinearDecay

learning_rate: 0.0002

start_epoch: 100

decay_epochs: 100

# will get from real dataset

iters_per_epoch: 1

optimizer:

optimizer_G:

name: Adam

net_names:

- netG

beta1: 0.5

optimizer_D:

name: Adam

net_names:

- netD

beta1: 0.5

log_config:

interval: 100

visiual_interval: 100

snapshot_config:

interval: 5