References

官方git：https://github.com/znxlwm/UGATIT-pytorch.git

paper：https://arxiv.org/abs/1907.10830

Machine

Tesla V100 [32G]

创建conda环境 Create a conda environment

Requirements：

• python == 3.6
• tensorflow == 1.14

conda create -n ugatit python = 3.6 tensorflow-gpu=1.14 tensorflow=1.14
conda activate ugatit
conda install opencv

Get U-GAT-IT

• get U-GAT-IT

git clone [<https://github.com/znxlwm/UGATIT-pytorch.git>](<https://github.com/znxlwm/UGATIT-pytorch.git>)

• get human face data(selfie2anime dataset) from "https://drive.google.com/file/d/1xOWj1UVgp6NKMT3HbPhBbtq2A4EDkghF/view"
• upload the testA/ and trainA/ dataset (the original human face data)

Preparations

• Data Preprocessing: Rename, Resize (with python scripts)
• Upload + Divide into testB/ and trainB/.