Minimal Convolutional Neural Networks for Temporal Anti Aliasing

Simple item page
dc.contributor.authorHerveau, Killianen_US
dc.contributor.authorPiochowiak, Maxen_US
dc.contributor.authorDachsbacher, Carstenen_US
dc.contributor.editorBikker, Jaccoen_US
dc.contributor.editorGribble, Christiaanen_US
dc.date.accessioned2023-06-25T09:06:52Z
dc.date.available2023-06-25T09:06:52Z
dc.date.issued2023
dc.description.abstractExisting deep learning methods for performing temporal anti aliasing (TAA) in rendering are either closed source or rely on upsampling networks with a large operation count which are expensive to evaluate. We propose a simple deep learning architecture for TAA combining only a few common primitives, easy to assemble and to change for application needs. We use a fully-convolutional neural network architecture with recurrent temporal feedback, motion vectors and depth values as input and show that a simple network can produce satisfactory results. Our architecture template, for which we provide code, introduces a method that adapts to different temporal subpixel offsets for accumulation without increasing the operation count. To this end, convolutional layers cycle through a set of different weights per temporal subpixel offset while their operations remain fixed. We analyze the effect of this method on image quality and present different tradeoffs for adapting the architecture. We show that our simple network performs remarkably better than variance clipping TAA, eliminating both flickering and ghosting without performing upsampling.en_US
dc.description.sectionheadersDeep Learning for Graphics
dc.description.seriesinformationHigh-Performance Graphics - Symposium Papers
dc.identifier.doi10.2312/hpg.20231134
dc.identifier.isbn978-3-03868-229-5
dc.identifier.issn2079-8687
dc.identifier.pages33-41
dc.identifier.pages9 pages
dc.identifier.urihttps://doi.org/10.2312/hpg.20231134
dc.identifier.urihttps://diglib.eg.org:443/handle/10.2312/hpg20231134
dc.publisherThe Eurographics Associationen_US
dc.rightsAttribution 4.0 International License
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/
dc.subjectCCS Concepts: Computing methodologies -> Antialiasing; Neural networks; Rendering
dc.subjectComputing methodologies
dc.subjectAntialiasing
dc.subjectNeural networks
dc.subjectRendering
dc.titleMinimal Convolutional Neural Networks for Temporal Anti Aliasingen_US
Files
Original bundle
Now showing 1 - 4 of 4
No Thumbnail Available
Name:
033-041.pdf
Size:
5.01 MB
Format:
Adobe Portable Document Format
Download
No Thumbnail Available
Name:
cafeteria.mp4
Size:
10.85 MB
Format:
Unknown data format
Download
No Thumbnail Available
Name:
livingroom.mp4
Size:
13.94 MB
Format:
Unknown data format
Download
No Thumbnail Available
Name:
stairs.mp4
Size:
71.03 MB
Format:
Unknown data format
Download
Collections
High-Performance Graphics 2023 - Symposium Papers