Thinking Fast and Slow: Efficient Text-to-Visual Retrieval with Transformers

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21730%2F21%3A00356140" target="_blank" >RIV/68407700:21730/21:00356140 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1109/CVPR46437.2021.00970" target="_blank" >https://doi.org/10.1109/CVPR46437.2021.00970</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/CVPR46437.2021.00970" target="_blank" >10.1109/CVPR46437.2021.00970</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Thinking Fast and Slow: Efficient Text-to-Visual Retrieval with Transformers

Popis výsledku v původním jazyce

Our objective is language-based search of large-scale image and video datasets. For this task, the approach that consists of independently mapping text and vision to a joint embedding space, a.k.a. dual encoders, is attractive as retrieval scales and is efficient for billions of images using approximate nearest neighbour search. An alternative approach of using vision-text transformers with cross-attention gives considerable improvements in accuracy over the joint embeddings, but is often inapplicable in practice for large-scale retrieval given the cost of the cross-attention mechanisms required for each sample at test time. This work combines the best of both worlds. We make the following three contributions. First, we equip transformerbased models with a new fine-grained cross-attention architecture, providing significant improvements in retrieval accuracy whilst preserving scalability. Second, we introduce a generic approach for combining a Fast dual encoder model with our Slow but accurate transformer-based model via distillation and re-ranking. Finally, we validate our approach on the Flickr30K image dataset where we show an increase in inference speed by several orders of magnitude while having results competitive to the state of the art. We also extend our method to the video domain, improving the state of the art on the VATEX dataset.

Název v anglickém jazyce

Thinking Fast and Slow: Efficient Text-to-Visual Retrieval with Transformers

Popis výsledku anglicky

Our objective is language-based search of large-scale image and video datasets. For this task, the approach that consists of independently mapping text and vision to a joint embedding space, a.k.a. dual encoders, is attractive as retrieval scales and is efficient for billions of images using approximate nearest neighbour search. An alternative approach of using vision-text transformers with cross-attention gives considerable improvements in accuracy over the joint embeddings, but is often inapplicable in practice for large-scale retrieval given the cost of the cross-attention mechanisms required for each sample at test time. This work combines the best of both worlds. We make the following three contributions. First, we equip transformerbased models with a new fine-grained cross-attention architecture, providing significant improvements in retrieval accuracy whilst preserving scalability. Second, we introduce a generic approach for combining a Fast dual encoder model with our Slow but accurate transformer-based model via distillation and re-ranking. Finally, we validate our approach on the Flickr30K image dataset where we show an increase in inference speed by several orders of magnitude while having results competitive to the state of the art. We also extend our method to the video domain, improving the state of the art on the VATEX dataset.

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/EF15_003%2F0000468" target="_blank" >EF15_003/0000468: Inteligentní strojové vnímání</a><br>

Návaznosti

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

Rok uplatnění

2021

Kód důvěrnosti údajů

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

Název statě ve sborníku

Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

ISBN

978-1-6654-4509-2

ISSN

1063-6919

e-ISSN

2575-7075

Počet stran výsledku

11

Strana od-do

9821-9831

Název nakladatele

IEEE Computer Society

Místo vydání

USA

Místo konání akce

Nashville

Datum konání akce

20. 6. 2021

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000742075007087