search

HsinYingLee / DRIT

Learning diverse image-to-image translation from unpaired data
844 stars 155 forks source link

Difference with MUNIT? #26

Closed Cuiyirui closed 5 years ago

Cuiyirui commented 5 years ago

The idea of this paper same to MUNIT https://github.com/NVlabs/MUNIT Both propose share content space and separate style space. I want to know the differences.

HsinYingLee commented 5 years ago

MUNIT and DRIT are concurrent works sharing a similar high-level idea. The key differences lie in the way they combine content and attribute features. MUNIT adopts AdaIn to transform content features based on attribute features, while DRIT provides two options: 1. For color-variation translation, simple concatenation works well. 2. For shape-variation translation, we adopt element-wise feature transformation.

Cuiyirui commented 5 years ago

MUNIT and DRIT are concurrent works sharing a similar high-level idea. The key differences lie in the way they combine content and attribute features. MUNIT adopts AdaIn to transform content features based on attribute features, while DRIT provides two options: 1. For color-variation translation, simple concatenation works well. 2. For shape-variation translation, we adopt element-wise feature transformation.

Thanks, I understood it.