DADO: A Depth-Attention Framework for Object Discovery

Federico Gonzalez; Estefania Talavera; Petia Radeva

doi:10.1007/978-3-032-05060-1_24

DADO: A Depth-Attention Framework for Object Discovery

Federico Gonzalez
, Estefania Talavera
, Petia Radeva

Datamanagement & Biometrics

Research output: Chapter in Book/Report/Conference proceeding › Chapter › Academic › peer-review

Abstract

Unsupervised object discovery, the task of identifying and localizing objects in images without human-annotated labels, remains a significant challenge and a growing focus in computer vision. In this work, we introduce a novel model, DADO (Depth-Attention self-supervised technique for Discovering unseen Objects), which combines an attention mechanism and a depth model to identify potential objects in images. To address challenges such as noisy attention maps or complex scenes with varying depth planes, DADO employs dynamic weighting to adaptively emphasize attention or depth features based on the global characteristics of each image. We evaluated DADO on standard benchmarks, where it outperforms state-of-the-art methods in object discovery accuracy and robustness without the need for fine-tuning.

Original language	English
Title of host publication	Computer Analysis of Images and Patterns
Subtitle of host publication	21st International Conference, CAIP 2025, Las Palmas de Gran Canaria, Spain, September 22–25, 2025, Proceedings, Part II
Editors	Modesto Castrillón-Santana, Carlos M. Travieso-González, David Freire-Obregón, Daniel Hernández-Sosa, Javier Lorenzo-Navarro, Oliverio J. Santana, Oscar Deniz Suarez
Publisher	Springer
Pages	281-291
Number of pages	11
ISBN (Electronic)	978-3-032-05060-1
ISBN (Print)	978-3-032-05059-5
DOIs	https://doi.org/10.1007/978-3-032-05060-1_24
Publication status	Published - 2026
Event	21st International Conference on Computer Analysis of Images and Patterns, CAIP 2025 - Museo Elder de la Ciencia y la Tecnología, Las Palmas de Gran Canaria, Spain Duration: 22 Sept 2025 → 25 Sept 2025 Conference number: 21 https://caip2025.com/

Publication series

Name	Lecture Notes in Computer Science
Volume	15622

Conference

Conference	21st International Conference on Computer Analysis of Images and Patterns, CAIP 2025
Abbreviated title	CAIP 2025
Country/Territory	Spain
City	Las Palmas de Gran Canaria
Period	22/09/25 → 25/09/25
Internet address	https://caip2025.com/

Keywords

NLA

Access to Document

10.1007/978-3-032-05060-1_24

Cite this

Gonzalez, F., Talavera, E., & Radeva, P. (2026). DADO: A Depth-Attention Framework for Object Discovery. In M. Castrillón-Santana, C. M. Travieso-González, D. Freire-Obregón, D. Hernández-Sosa, J. Lorenzo-Navarro, O. J. Santana, & O. Deniz Suarez (Eds.), Computer Analysis of Images and Patterns: 21st International Conference, CAIP 2025, Las Palmas de Gran Canaria, Spain, September 22–25, 2025, Proceedings, Part II (pp. 281-291). ( Lecture Notes in Computer Science; Vol. 15622). Springer. https://doi.org/10.1007/978-3-032-05060-1_24

Gonzalez, Federico ; Talavera, Estefania ; Radeva, Petia. / DADO: A Depth-Attention Framework for Object Discovery. Computer Analysis of Images and Patterns: 21st International Conference, CAIP 2025, Las Palmas de Gran Canaria, Spain, September 22–25, 2025, Proceedings, Part II. editor / Modesto Castrillón-Santana ; Carlos M. Travieso-González ; David Freire-Obregón ; Daniel Hernández-Sosa ; Javier Lorenzo-Navarro ; Oliverio J. Santana ; Oscar Deniz Suarez. Springer, 2026. pp. 281-291 ( Lecture Notes in Computer Science).

@inbook{5dfbb70d65e0498796eedc2c156676f3,

title = "DADO: A Depth-Attention Framework for Object Discovery",

abstract = "Unsupervised object discovery, the task of identifying and localizing objects in images without human-annotated labels, remains a significant challenge and a growing focus in computer vision. In this work, we introduce a novel model, DADO (Depth-Attention self-supervised technique for Discovering unseen Objects), which combines an attention mechanism and a depth model to identify potential objects in images. To address challenges such as noisy attention maps or complex scenes with varying depth planes, DADO employs dynamic weighting to adaptively emphasize attention or depth features based on the global characteristics of each image. We evaluated DADO on standard benchmarks, where it outperforms state-of-the-art methods in object discovery accuracy and robustness without the need for fine-tuning.",

keywords = "NLA",

author = "Federico Gonzalez and Estefania Talavera and Petia Radeva",

year = "2026",

doi = "10.1007/978-3-032-05060-1\_24",

language = "English",

isbn = "978-3-032-05059-5",

series = " Lecture Notes in Computer Science",

publisher = "Springer",

pages = "281--291",

editor = "Modesto Castrill{\'o}n-Santana and Travieso-Gonz{\'a}lez, \{Carlos M.\} and David Freire-Obreg{\'o}n and Daniel Hern{\'a}ndez-Sosa and Javier Lorenzo-Navarro and Santana, \{Oliverio J.\} and \{Deniz Suarez\}, Oscar",

booktitle = "Computer Analysis of Images and Patterns",

address = "Germany",

note = "21st International Conference on Computer Analysis of Images and Patterns, CAIP 2025, CAIP 2025 ; Conference date: 22-09-2025 Through 25-09-2025",

url = "https://caip2025.com/",

}

Gonzalez, F, Talavera, E & Radeva, P 2026, DADO: A Depth-Attention Framework for Object Discovery. in M Castrillón-Santana, CM Travieso-González, D Freire-Obregón, D Hernández-Sosa, J Lorenzo-Navarro, OJ Santana & O Deniz Suarez (eds), Computer Analysis of Images and Patterns: 21st International Conference, CAIP 2025, Las Palmas de Gran Canaria, Spain, September 22–25, 2025, Proceedings, Part II. Lecture Notes in Computer Science, vol. 15622, Springer, pp. 281-291, 21st International Conference on Computer Analysis of Images and Patterns, CAIP 2025, Las Palmas de Gran Canaria, Spain, 22/09/25. https://doi.org/10.1007/978-3-032-05060-1_24

DADO: A Depth-Attention Framework for Object Discovery. / Gonzalez, Federico; Talavera, Estefania; Radeva, Petia.
Computer Analysis of Images and Patterns: 21st International Conference, CAIP 2025, Las Palmas de Gran Canaria, Spain, September 22–25, 2025, Proceedings, Part II. ed. / Modesto Castrillón-Santana; Carlos M. Travieso-González; David Freire-Obregón; Daniel Hernández-Sosa; Javier Lorenzo-Navarro; Oliverio J. Santana; Oscar Deniz Suarez. Springer, 2026. p. 281-291 ( Lecture Notes in Computer Science; Vol. 15622).

Research output: Chapter in Book/Report/Conference proceeding › Chapter › Academic › peer-review

TY - CHAP

T1 - DADO: A Depth-Attention Framework for Object Discovery

AU - Gonzalez, Federico

AU - Talavera, Estefania

AU - Radeva, Petia

N1 - Conference code: 21

PY - 2026

Y1 - 2026

N2 - Unsupervised object discovery, the task of identifying and localizing objects in images without human-annotated labels, remains a significant challenge and a growing focus in computer vision. In this work, we introduce a novel model, DADO (Depth-Attention self-supervised technique for Discovering unseen Objects), which combines an attention mechanism and a depth model to identify potential objects in images. To address challenges such as noisy attention maps or complex scenes with varying depth planes, DADO employs dynamic weighting to adaptively emphasize attention or depth features based on the global characteristics of each image. We evaluated DADO on standard benchmarks, where it outperforms state-of-the-art methods in object discovery accuracy and robustness without the need for fine-tuning.

AB - Unsupervised object discovery, the task of identifying and localizing objects in images without human-annotated labels, remains a significant challenge and a growing focus in computer vision. In this work, we introduce a novel model, DADO (Depth-Attention self-supervised technique for Discovering unseen Objects), which combines an attention mechanism and a depth model to identify potential objects in images. To address challenges such as noisy attention maps or complex scenes with varying depth planes, DADO employs dynamic weighting to adaptively emphasize attention or depth features based on the global characteristics of each image. We evaluated DADO on standard benchmarks, where it outperforms state-of-the-art methods in object discovery accuracy and robustness without the need for fine-tuning.

KW - NLA

UR - https://www.scopus.com/pages/publications/105024538375

U2 - 10.1007/978-3-032-05060-1_24

DO - 10.1007/978-3-032-05060-1_24

M3 - Chapter

SN - 978-3-032-05059-5

T3 - Lecture Notes in Computer Science

SP - 281

EP - 291

BT - Computer Analysis of Images and Patterns

A2 - Castrillón-Santana, Modesto

A2 - Travieso-González, Carlos M.

A2 - Freire-Obregón, David

A2 - Hernández-Sosa, Daniel

A2 - Lorenzo-Navarro, Javier

A2 - Santana, Oliverio J.

A2 - Deniz Suarez, Oscar

PB - Springer

T2 - 21st International Conference on Computer Analysis of Images and Patterns, CAIP 2025

Y2 - 22 September 2025 through 25 September 2025

ER -

Gonzalez F, Talavera E, Radeva P. DADO: A Depth-Attention Framework for Object Discovery. In Castrillón-Santana M, Travieso-González CM, Freire-Obregón D, Hernández-Sosa D, Lorenzo-Navarro J, Santana OJ, Deniz Suarez O, editors, Computer Analysis of Images and Patterns: 21st International Conference, CAIP 2025, Las Palmas de Gran Canaria, Spain, September 22–25, 2025, Proceedings, Part II. Springer. 2026. p. 281-291. ( Lecture Notes in Computer Science). Epub 2025 Sept 17. doi: 10.1007/978-3-032-05060-1_24