ALMNet: Adjacent Layer Driven Multiscale Features for Salient Object Detection

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F62690094%3A18450%2F21%3A50018317" target="_blank" >RIV/62690094:18450/21:50018317 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/9524698" target="_blank" >https://ieeexplore.ieee.org/document/9524698</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/TIM.2021.3108503" target="_blank" >10.1109/TIM.2021.3108503</a>

Jazyk výsledku

angličtina

Název v původním jazyce

ALMNet: Adjacent Layer Driven Multiscale Features for Salient Object Detection

Popis výsledku v původním jazyce

Nowadays, the usage of deep learning-based approaches for salient object detection (SOD) is increasing exponentially to detect and localize visually distinct regions in static images. However, the variability in scales of salient objects requires further attention given the abstract nature of the multi-layer feature hierarchy of convolution neural networks (CNNs). Firstly, feature maps of different layers in CNNs embed abstract information about objects that changes with the object’s scale. Secondly, the progressive feature fusion in models such as the feature pyramid network loses its effectiveness in detecting sharp boundaries due to the late fusion of detailed features. This work proposes two modules namely, adjacent layer attention block and partial encoder-decoder block to handle the aforementioned issues. The proposed adjacent layer attention block facilitates communication among the layers of closest abstraction to mine abundant scale features at the current resolution. The resultant integrated feature at a resolution contains detailed as well as semantic information from interaction among the adjacent layers useful to extract scale information of complex objects. A partial encoder-decoder module utilizes the resolution-specific integrated features from the adjacent layer attention block of its corresponding encoder to generate multi-scale features, and fuse them in an top-down manner. This level-specific distribution of aggregated features within a partial encoder-decoder helps coarser-layers of the network to acquire boundary information. Experimental results on five broadly used salient object detection datasets are compared with recent 20 state-of-the-art SOD models. The proposed method performs favorably against its competitors without any pre-processing or post-processing. IEEE

Název v anglickém jazyce

ALMNet: Adjacent Layer Driven Multiscale Features for Salient Object Detection

Popis výsledku anglicky

Nowadays, the usage of deep learning-based approaches for salient object detection (SOD) is increasing exponentially to detect and localize visually distinct regions in static images. However, the variability in scales of salient objects requires further attention given the abstract nature of the multi-layer feature hierarchy of convolution neural networks (CNNs). Firstly, feature maps of different layers in CNNs embed abstract information about objects that changes with the object’s scale. Secondly, the progressive feature fusion in models such as the feature pyramid network loses its effectiveness in detecting sharp boundaries due to the late fusion of detailed features. This work proposes two modules namely, adjacent layer attention block and partial encoder-decoder block to handle the aforementioned issues. The proposed adjacent layer attention block facilitates communication among the layers of closest abstraction to mine abundant scale features at the current resolution. The resultant integrated feature at a resolution contains detailed as well as semantic information from interaction among the adjacent layers useful to extract scale information of complex objects. A partial encoder-decoder module utilizes the resolution-specific integrated features from the adjacent layer attention block of its corresponding encoder to generate multi-scale features, and fuse them in an top-down manner. This level-specific distribution of aggregated features within a partial encoder-decoder helps coarser-layers of the network to acquire boundary information. Experimental results on five broadly used salient object detection datasets are compared with recent 20 state-of-the-art SOD models. The proposed method performs favorably against its competitors without any pre-processing or post-processing. IEEE

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

20201 - Electrical and electronic engineering

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2021

Kód důvěrnosti údajů

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

Název periodika

IEEE Transactions on Instrumentation and Measurement

ISSN

0018-9456

e-ISSN

—

Svazek periodika

70

Číslo periodika v rámci svazku

August

Stát vydavatele periodika

US - Spojené státy americké

Počet stran výsledku

14

Strana od-do

"Article Number: 2513214"

Kód UT WoS článku

000698642300004

EID výsledku v databázi Scopus

2-s2.0-85114729199