EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision Applications

Muhammad Maaz*, Abdelrahman Shaker, Hisham Cholakkal, Salman Khan, Syed Waqas Zamir, Rao Muhammad Anwer, Fahad Shahbaz Khan

*Tämän työn vastaava kirjoittaja

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussaConference article in proceedingsScientificvertaisarvioitu

22 Sitaatiot (Scopus)

Abstrakti

In the pursuit of achieving ever-increasing accuracy, large and complex neural networks are usually developed. Such models demand high computational resources and therefore cannot be deployed on edge devices. It is of great interest to build resource-efficient general purpose networks due to their usefulness in several application areas. In this work, we strive to effectively combine the strengths of both CNN and Transformer models and propose a new efficient hybrid architecture EdgeNeXt. Specifically in EdgeNeXt, we introduce split depth-wise transpose attention (STDA) encoder that splits input tensors into multiple channel groups and utilizes depth-wise convolution along with self-attention across channel dimensions to implicitly increase the receptive field and encode multi-scale features. Our extensive experiments on classification, detection and segmentation tasks, reveal the merits of the proposed approach, outperforming state-of-the-art methods with comparatively lower compute requirements. Our EdgeNeXt model with 1.3M parameters achieves 71.2% top-1 accuracy on ImageNet-1K, outperforming MobileViT with an absolute gain of 2.2% with 28% reduction in FLOPs. Further, our EdgeNeXt model with 5.6M parameters achieves 79.4% top-1 accuracy on ImageNet-1K. The code and models are available at https://t.ly/_Vu9.

AlkuperäiskieliEnglanti
OtsikkoComputer Vision – ECCV 2022 Workshops
AlaotsikkoTel Aviv, Israel, October 23–27, 2022, Proceedings, Part VII
ToimittajatLeonid Karlinsky, Tomer Michaeli, Ko Nishino
KustantajaSpringer
Sivut3-20
Sivumäärä18
ISBN (painettu)978-3-031-25081-1
DOI - pysyväislinkit
TilaJulkaistu - 2023
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisussa
TapahtumaEuropean Conference on Computer Vision - Tel Aviv, Israel
Kesto: 23 lokak. 202227 lokak. 2022
Konferenssinumero: 17
https://eccv2022.ecva.net

Julkaisusarja

NimiLecture Notes in Computer Science
Vuosikerta13807 LNCS
ISSN (painettu)0302-9743
ISSN (elektroninen)1611-3349

Conference

ConferenceEuropean Conference on Computer Vision
LyhennettäECCV
Maa/AlueIsrael
KaupunkiTel Aviv
Ajanjakso23/10/202227/10/2022
www-osoite

Sormenjälki

Sukella tutkimusaiheisiin 'EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision Applications'. Ne muodostavat yhdessä ainutlaatuisen sormenjäljen.

Siteeraa tätä