The goal of I2I(image-to-image) is to convert an input image $x_A$ frim the socurce domain $A$ to a target domain $B$ with the intrinsic source content preserved and the extrinsic target style transferred

Untitled

1. Backbone of I2I

1-1. Variational AutoEncoder

1-2. Generative Adversarial Networks

1-3. Evaluation Metrics

2. Two-Domain Image-To-Image Translation

Untitled

2-1. Supervised I2I Translation

2-1-1. Single-modal Output

2-1-2. Multi-modal Output

2-2. Unsupervised I2I Translation (UI2I)

2-2-1. Single-modal Output

  1. Translation using Cycle-consistency Constraint
  2. Translation beyond Cycle-consistency Constraint
  3. Translation of Fine-grained Objects
  4. Translation by combining knowledge in other fields

2-2-2. Multi-modal Output