Composable Unpaired Image to Image Translation
This work addresses the problem of limited flexibility in image translation for computer vision researchers, offering a more scalable and composable approach, though it is incremental as it builds on existing methods.
The paper tackles the limitation of unpaired image-to-image translation being restricted to single distribution pairs by extending it to a scalable multi-distribution mechanism with composable models, enabling generation of images with unseen characteristics and showing improved sample quality through decoupled training.
There has been remarkable recent work in unpaired image-to-image translation. However, they're restricted to translation on single pairs of distributions, with some exceptions. In this study, we extend one of these works to a scalable multidistribution translation mechanism. Our translation models not only converts from one distribution to another but can be stacked to create composite translation functions. We show that this composite property makes it possible to generate images with characteristics not seen in the training set. We also propose a decoupled training mechanism to train multiple distributions separately, which we show, generates better samples than isolated joint training. Further, we do a qualitative and quantitative analysis to assess the plausibility of the samples. The code is made available at https://github.com/lgraesser/im2im2im.