Tunable Convolutions with Parametric Multi-Loss Optimization

Matteo Maggioni, Thomas Tanay, Francesca Babiloni, Steven McDonagh, Ales Leonardis

Research output: Chapter in Book/Report/Conference proceedingConference contribution

8 Downloads (Pure)

Abstract

Behavior of neural networks is irremediably determined by the specific loss and data used during training. However it is often desirable to tune the model at inference time based on external factors such as preferences of the user or dynamic characteristics of the data. This is especially important to balance the perception-distortion trade-off of ill-posed image-to-image translation tasks. In this work, we propose to optimize a parametric tunable convolutional layer, which includes a number of different kernels, using a parametric multi-loss, which includes an equal number of objectives. Our key insight is to use a shared set of parameters to dynamically interpolate both the objectives and the kernels. During training, these parameters are sampled at random to explicitly optimize all possible combinations of objectives and consequently disentangle their effect into the corresponding kernels. During inference, these parameters become interactive inputs of the model hence enabling reliable and consistent control over the model behavior. Extensive experimental results demonstrate that our tunable convolutions effectively work as a drop-in replacement for traditional convolutions in existing neural networks at virtually no extra computational cost, outperforming state-of-the-art control strategies in a wide range of applications; including image denoising, deblurring, super-resolution, and style transfer.
Original languageEnglish
Title of host publication2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
PublisherIEEE
Pages20226-20236
Number of pages11
Publication statusPublished - 22 Jun 2023
Event34th IEEE/CVF Conference on Computer Vision and Pattern Recognition - Vancouver, Canada
Duration: 18 Jun 202322 Jun 2023

Publication series

NameProceedings. IEEE Computer Society Conference on Computer Vision and Pattern Recognition
PublisherIEEE
ISSN (Print)1063-6919
ISSN (Electronic)2575-7075

Conference

Conference34th IEEE/CVF Conference on Computer Vision and Pattern Recognition
Country/TerritoryCanada
CityVancouver
Period18/06/2322/06/23

Fingerprint

Dive into the research topics of 'Tunable Convolutions with Parametric Multi-Loss Optimization'. Together they form a unique fingerprint.

Cite this