songVAE

The 1,000 training images output by songMix were trained for a total of 80,000 steps, and the model was further refined.

The emphasis is on depicting the skin color of people and correcting blown-out highlights in the background.

~

songMixで出力した教師画像1000枚を総計80000stepで学習させ、調整を重ねました。

人物の肌の色の描写と背景の白とび修正に重点を置いています。