Preface

Based on the models trained previously with Dunhuang cave painting (not face).

stylegan2使用敦煌壁画作为训练集训练

Generate-Basic

使用的network是根据壁画图像(1515张)训练的最新模型文件: 00001-2-dunhuang-transferred-mirror-auto1-bg-resumecustom/network-snapshot-000080.pkl

1. Generate curated images without truncation

python generate.py --outdir=./out --trunc=1 --seeds=85,265,297,849 --network=./training-runs/00001-2-dunhuang-transferred-mirror-auto1-bg-resumecustom/network-snapshot-000080.pkl

seed0085

seed0849

seed0297

seed0265

2.Generate uncurated images with truncation

python generate.py --outdir=./out --trunc=0.7 --seeds=600-605 --network=./training-runs/00001-2-dunhuang-transferred-mirror-auto1-bg-resumecustom/network-snapshot-000080.pkl

seed 0600 ~ seed 0605

3. Style mixing

python style_mixing.py --outdir=./out --rows=85,100,75,458,1500 --cols=55,821,1789,293 --network=./training-runs/00001-2-dunhuang-transferred-mirror-auto1-bg-resumecustom/network-snapshot-000080.pkl

• Error:【未解决】

4. Render an image from projected W