Pattern spotting in historical documents using convolutional models

Ignacio Úbeda; Jose M. Saavedra; Stéphane Nicolas; Caroline Petitjean; Laurent Heutte

doi:10.1145/3352631.3352645

Pattern spotting in historical documents using convolutional models

Ignacio Úbeda, Jose M. Saavedra, Stéphane Nicolas, Caroline Petitjean, Laurent Heutte

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

4 Scopus citations

Abstract

Pattern spotting consists of searching in a collection of historical document images for occurrences of a graphical object using an image query. Contrary to object detection, no prior information nor predefined class is given about the query so training a model of the object is not feasible. In this paper, a convolutional neural network approach is proposed to tackle this problem. We use RetinaNet as a feature extractor to obtain multiscale embeddings of the regions of the documents and also for the queries. Experiments conducted on the DocExplore dataset show that our proposal is better at locating patterns and requires less storage for indexing images than the state-of-the-art system, but fails at retrieving multiple pages containing instances of the query.

Original language	English
Title of host publication	Proceedings of the 2019 Workshop on Historical Document Imaging and Processing, HIP 2019
Publisher	Association for Computing Machinery
Pages	60-65
Number of pages	6
ISBN (Electronic)	9781450376686
DOIs	https://doi.org/10.1145/3352631.3352645
State	Published - 20 Sep 2019
Externally published	Yes
Event	5th International Workshop on Historical Document Imaging and Processing, HIP 2019, held in conjunction with ICDAR 2019 - Sydney, Australia Duration: 20 Sep 2019 → 21 Sep 2019

Publication series

Name	ACM International Conference Proceeding Series

Conference

Conference	5th International Workshop on Historical Document Imaging and Processing, HIP 2019, held in conjunction with ICDAR 2019
Country/Territory	Australia
City	Sydney
Period	20/09/19 → 21/09/19

Bibliographical note

Publisher Copyright:
© 2019 Association for Computing Machinery.

Keywords

Convolutional neural network
Historical documents
Image retrieval
Pattern spotting

Access to Document

10.1145/3352631.3352645

Cite this

Úbeda, I., Saavedra, J. M., Nicolas, S., Petitjean, C., & Heutte, L. (2019). Pattern spotting in historical documents using convolutional models. In Proceedings of the 2019 Workshop on Historical Document Imaging and Processing, HIP 2019 (pp. 60-65). (ACM International Conference Proceeding Series). Association for Computing Machinery. https://doi.org/10.1145/3352631.3352645

@inproceedings{8b205c82942d4f5798dab97206c78719,

title = "Pattern spotting in historical documents using convolutional models",

abstract = "Pattern spotting consists of searching in a collection of historical document images for occurrences of a graphical object using an image query. Contrary to object detection, no prior information nor predefined class is given about the query so training a model of the object is not feasible. In this paper, a convolutional neural network approach is proposed to tackle this problem. We use RetinaNet as a feature extractor to obtain multiscale embeddings of the regions of the documents and also for the queries. Experiments conducted on the DocExplore dataset show that our proposal is better at locating patterns and requires less storage for indexing images than the state-of-the-art system, but fails at retrieving multiple pages containing instances of the query.",

keywords = "Convolutional neural network, Historical documents, Image retrieval, Pattern spotting",

author = "Ignacio {\'U}beda and Saavedra, {Jose M.} and St{\'e}phane Nicolas and Caroline Petitjean and Laurent Heutte",

note = "Publisher Copyright: {\textcopyright} 2019 Association for Computing Machinery.; 5th International Workshop on Historical Document Imaging and Processing, HIP 2019, held in conjunction with ICDAR 2019 ; Conference date: 20-09-2019 Through 21-09-2019",

year = "2019",

month = sep,

day = "20",

doi = "10.1145/3352631.3352645",

language = "English",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "60--65",

booktitle = "Proceedings of the 2019 Workshop on Historical Document Imaging and Processing, HIP 2019",

}

Úbeda, I, Saavedra, JM, Nicolas, S, Petitjean, C & Heutte, L 2019, Pattern spotting in historical documents using convolutional models. in Proceedings of the 2019 Workshop on Historical Document Imaging and Processing, HIP 2019. ACM International Conference Proceeding Series, Association for Computing Machinery, pp. 60-65, 5th International Workshop on Historical Document Imaging and Processing, HIP 2019, held in conjunction with ICDAR 2019, Sydney, Australia, 20/09/19. https://doi.org/10.1145/3352631.3352645

Pattern spotting in historical documents using convolutional models. / Úbeda, Ignacio; Saavedra, Jose M.; Nicolas, Stéphane et al.
Proceedings of the 2019 Workshop on Historical Document Imaging and Processing, HIP 2019. Association for Computing Machinery, 2019. p. 60-65 (ACM International Conference Proceeding Series).

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

TY - GEN

T1 - Pattern spotting in historical documents using convolutional models

AU - Úbeda, Ignacio

AU - Saavedra, Jose M.

AU - Nicolas, Stéphane

AU - Petitjean, Caroline

AU - Heutte, Laurent

PY - 2019/9/20

Y1 - 2019/9/20

N2 - Pattern spotting consists of searching in a collection of historical document images for occurrences of a graphical object using an image query. Contrary to object detection, no prior information nor predefined class is given about the query so training a model of the object is not feasible. In this paper, a convolutional neural network approach is proposed to tackle this problem. We use RetinaNet as a feature extractor to obtain multiscale embeddings of the regions of the documents and also for the queries. Experiments conducted on the DocExplore dataset show that our proposal is better at locating patterns and requires less storage for indexing images than the state-of-the-art system, but fails at retrieving multiple pages containing instances of the query.

AB - Pattern spotting consists of searching in a collection of historical document images for occurrences of a graphical object using an image query. Contrary to object detection, no prior information nor predefined class is given about the query so training a model of the object is not feasible. In this paper, a convolutional neural network approach is proposed to tackle this problem. We use RetinaNet as a feature extractor to obtain multiscale embeddings of the regions of the documents and also for the queries. Experiments conducted on the DocExplore dataset show that our proposal is better at locating patterns and requires less storage for indexing images than the state-of-the-art system, but fails at retrieving multiple pages containing instances of the query.

KW - Convolutional neural network

KW - Historical documents

KW - Image retrieval

KW - Pattern spotting

UR - http://www.scopus.com/inward/record.url?scp=85074773138&partnerID=8YFLogxK

U2 - 10.1145/3352631.3352645

DO - 10.1145/3352631.3352645

M3 - Conference contribution

AN - SCOPUS:85074773138

T3 - ACM International Conference Proceeding Series

SP - 60

EP - 65

BT - Proceedings of the 2019 Workshop on Historical Document Imaging and Processing, HIP 2019

PB - Association for Computing Machinery

T2 - 5th International Workshop on Historical Document Imaging and Processing, HIP 2019, held in conjunction with ICDAR 2019

Y2 - 20 September 2019 through 21 September 2019

ER -

Pattern spotting in historical documents using convolutional models

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this