Papers › Powers of layers for image-to-image translation

Powers of layers for image-to-image translation

13 Aug 2020arXiv:2008.05763archive 2025-07-28

Hugo Touvron, Matthijs Douze, Matthieu Cord, Hervé Jégou

We propose a simple architecture to address unpaired image-to-image translation tasks: style or class transfer, denoising, deblurring, deblocking, etc. We start from an image autoencoder architecture with fixed weights. For each task we learn a residual block operating in the latent space, which is iteratively called until the target domain is reached. A specific training schedule is required to alleviate the exponentiation effect of the iterations. At test time, it offers several advantages: the number of weight parameters is limited and the compositional design allows one to modulate the strength of the transformation with the number of iterations. This is useful, for instance, when the type or amount of noise to suppress is not known in advance. Experimentally, we provide proofs of concepts showing the interest of our method for many transformations. The performance of our model is comparable or better than CycleGAN with significantly fewer parameters.

PaperPDFConference PDF

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

DeblurringDenoisingImage-to-Image TranslationTranslation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image-to-Image Translation horse2zebra PoL (CycleGAN) Frechet Inception Distance 53.0 #3 of 4 Archive leaderboard report
Image-to-Image Translation horse2zebra PoL (CycleGAN) Number of params 15.9M #3 of 4 Archive leaderboard report
Image-to-Image Translation photo2vangogh PoL (CycleGAN) Frechet Inception Distance 152.7 #3 of 3 Archive leaderboard report
Image-to-Image Translation photo2vangogh PoL (CycleGAN) Number of params 15.9M #3 of 3 Archive leaderboard report
Image-to-Image Translation vangogh2photo PoL (CycleGAN) Frechet Inception Distance 134.4 #2 of 3 Archive leaderboard report
Image-to-Image Translation vangogh2photo PoL (CycleGAN) Number of Params 15.9M #2 of 3 Archive leaderboard report
Image-to-Image Translation zebra2horse PoL (CycleGAN) Frechet Inception Distance 112.3 #3 of 3 Archive leaderboard report
Image-to-Image Translation zebra2horse PoL (CycleGAN) Number of params 15.9M #3 of 3 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Batch NormalizationConvolutionCycle Consistency LossGAN Least Squares LossInstance NormalizationPatchGANReLUResidual BlockResidual ConnectionSigmoid ActivationTanh Activation

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections