Single reference

Deepremaster

Paper: https://github.com/satoshiiizuka/siggraphasia2019_remastering

This model is originally designed for film colorization. To run this benchmark the input image is duplicated 5 times. The reference images are supposed to be colored frames chosen from the movies.

This means that significant differences in the reference images cannot be used, as illustrated below. (‘Recolor source’ vs other rows)

Task	Image #1	Image #2	Image #3	Image #4
Recolor source
Inverted reference
Task	Image #1	Image #2	Image #3	Reference
Full correspondence
Full correspondence
Partial reference
Partial reference
Partial source
Partial source
Semantic correspondence strong
Semantic correspondence strong
Semantic correspondence weak
Semantic correspondence weak
Distractors
Random noise
Random noise
Gray
Gray
Contemporary
Contemporary
Contemporary