Large-to-small Image Resolution Asymmetry in Deep Metric Learning

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F23%3A00362078" target="_blank" >RIV/68407700:21230/23:00362078 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1109/WACV56688.2023.00150" target="_blank" >https://doi.org/10.1109/WACV56688.2023.00150</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/WACV56688.2023.00150" target="_blank" >10.1109/WACV56688.2023.00150</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Large-to-small Image Resolution Asymmetry in Deep Metric Learning

Popis výsledku v původním jazyce

Deep metric learning for vision is trained by optimizing a representation network to map (non-)matching image pairs to (non-)similar representations. During testing, which typically corresponds to image retrieval, both database and query examples are processed by the same network to obtain the representation used for similarity estimation and ranking. In this work, we explore an asymmetric setup by light-weight processing of the query at a small image resolution to enable fast representation extraction. The goal is to obtain a network for database examples that is trained to operate on large resolution images and benefits from fine-grained image details, and a second network for query examples that operates on small resolution images but preserves a representation space aligned with that of the database network. We achieve this with a distillation approach that transfers knowledge from a fixed teacher network to a student via a loss that operates per image and solely relies on coupled augmentations without the use of any labels. In contrast to prior work that explores such asymmetry from the point of view of different network architectures, this work uses the same architecture but modifies the image resolution. We conclude that resolution asymmetry is a better way to optimize the performance/efficiency trade-off than architecture asymmetry. Evaluation is performed on three standard deep metric learning benchmarks, namely CUB200, Cars196, and SOP. Code: https://github.com/pavelsuma/raml

Název v anglickém jazyce

Large-to-small Image Resolution Asymmetry in Deep Metric Learning

Popis výsledku anglicky

Deep metric learning for vision is trained by optimizing a representation network to map (non-)matching image pairs to (non-)similar representations. During testing, which typically corresponds to image retrieval, both database and query examples are processed by the same network to obtain the representation used for similarity estimation and ranking. In this work, we explore an asymmetric setup by light-weight processing of the query at a small image resolution to enable fast representation extraction. The goal is to obtain a network for database examples that is trained to operate on large resolution images and benefits from fine-grained image details, and a second network for query examples that operates on small resolution images but preserves a representation space aligned with that of the database network. We achieve this with a distillation approach that transfers knowledge from a fixed teacher network to a student via a loss that operates per image and solely relies on coupled augmentations without the use of any labels. In contrast to prior work that explores such asymmetry from the point of view of different network architectures, this work uses the same architecture but modifies the image resolution. We conclude that resolution asymmetry is a better way to optimize the performance/efficiency trade-off than architecture asymmetry. Evaluation is performed on three standard deep metric learning benchmarks, namely CUB200, Cars196, and SOP. Code: https://github.com/pavelsuma/raml

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/GM21-28830M" target="_blank" >GM21-28830M: Učení Univerzální Vizuální Reprezentace s Omezenou Supervizí</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proc. of the 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

ISBN

978-1-6654-9346-8

ISSN

2472-6737

e-ISSN

2642-9381

Počet stran výsledku

10

Strana od-do

1451-1460

Název nakladatele

IEEE

Místo vydání

Piscataway

Místo konání akce

Waikoloa

Datum konání akce

3. 1. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000971500201053