GATE: Geometry-Aware Trained Encoding

Simple item page
dc.contributor.authorBoksansky, Jakuben_US
dc.contributor.authorMeister, Danielen_US
dc.contributor.authorBenthin, Carstenen_US
dc.contributor.editorKnoll, Aaronen_US
dc.contributor.editorPeters, Christophen_US
dc.date.accessioned2025-06-20T07:27:05Z
dc.date.available2025-06-20T07:27:05Z
dc.date.issued2025
dc.description.abstractThe encoding of input parameters is one of the fundamental building blocks of neural network algorithms. Its goal is to map the input data to a higher-dimensional space [RBA*19], typically supported by trained feature vectors [MESK22]. The mapping is crucial for the efficiency and approximation quality of neural networks. We propose a novel geometry-aware encoding called GATE that stores feature vectors on the surface of triangular meshes. Our encoding is suitable for neural rendering-related algorithms, for example, neural radiance caching [MRNK21]. It also avoids limitations of previous hash-based encoding schemes, such as hash collisions, selection of resolution versus scene size, and divergent memory access. Our approach decouples feature vector density from geometry density using mesh colors [YKH10], while allowing for finer control over neural network training and adaptive level-of-detail.en_US
dc.description.sectionheadersNeural Textures and Encodings
dc.description.seriesinformationHigh-Performance Graphics - Symposium Papers
dc.identifier.doi10.2312/hpg.20251175
dc.identifier.isbn978-3-03868-291-2
dc.identifier.issn2079-8687
dc.identifier.pages9 pages
dc.identifier.urihttps://doi.org/10.2312/hpg.20251175
dc.identifier.urihttps://diglib.eg.org/handle/10.2312/hpg20251175
dc.publisherThe Eurographics Associationen_US
dc.rightsAttribution 4.0 International License
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/
dc.titleGATE: Geometry-Aware Trained Encodingen_US
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
hpg20251175.pdf
Size:
116.13 MB
Format:
Adobe Portable Document Format
Download
Collections
High-Performance Graphics 2025 - Symposium Papers