Gradient field descriptor for sketch based retrieval and localization

Rui Hu; Mark Barnard; John Collomosse

doi:10.1109/ICIP.2010.5649331

Gradient field descriptor for sketch based retrieval and localization

Rui Hu, Mark Barnard, John Collomosse

Research output: Contribution to conference › Paper › peer-review

Abstract

We present an image retrieval system driven by free-hand sketched queries depicting shape. We introduce Gradient Field HoG (GF-HOG) as a depiction invariant image descriptor, encapsulating local spatial structure in the sketch and facilitating efficient codebook based retrieval. We show improved retrieval accuracy over 3 leading descriptors (Self Similarity, SIFT, HoG) across two datasets (Flickr160, ETHZ extended objects), and explain how GF-HOG can be combined with RANSAC to localize sketched objects within relevant images. We also demonstrate a prototype sketch driven photo montage application based on our system.

Original language	English
DOIs	https://doi.org/10.1109/ICIP.2010.5649331
Publication status	Published - Sept 2010
Event	IEEE International Conference on Image Processing - Hong Kong, Hong Kong Duration: 12 Sept 2010 → 15 Sept 2010

Conference

Conference	IEEE International Conference on Image Processing
Period	12/09/10 → 15/09/10

Bibliographical note

Note: Published in: 2010 IEEE International Conference on Image Processing. Piscataway, NJ : Institute of Electrical and Electronics Engineers. ISSN 1522-4880 ISBN 9781424479924

Organising Body: IEEE

Keywords

Computer science and informatics

Access to Document

10.1109/ICIP.2010.5649331

https://doi.org/10.1109/ICIP.2010.5649331

1 Conference contribution

Gradient field descriptor for sketch based retrieval and localization
Hu, R., Barnard, M. & Collomosse, J., Sept 2010, Published in: 2010 IEEE International Conference on Image Processing. Piscataway, NJ : Institute of Electrical and Electronics Engineers. ISSN 1522-4880 ISBN 9781424479924 Organising Body: IEEE Organising Body: IEEE. p. 1025 - 1028
Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Cite this

@conference{a591245a74ab452ebe5c21df0ee405d9,

title = "Gradient field descriptor for sketch based retrieval and localization",

abstract = "We present an image retrieval system driven by free-hand sketched queries depicting shape. We introduce Gradient Field HoG (GF-HOG) as a depiction invariant image descriptor, encapsulating local spatial structure in the sketch and facilitating efficient codebook based retrieval. We show improved retrieval accuracy over 3 leading descriptors (Self Similarity, SIFT, HoG) across two datasets (Flickr160, ETHZ extended objects), and explain how GF-HOG can be combined with RANSAC to localize sketched objects within relevant images. We also demonstrate a prototype sketch driven photo montage application based on our system.",

keywords = "Computer science and informatics",

author = "Rui Hu and Mark Barnard and John Collomosse",

note = "Note: Published in: 2010 IEEE International Conference on Image Processing. Piscataway, NJ : Institute of Electrical and Electronics Engineers. ISSN 1522-4880 ISBN 9781424479924 Organising Body: IEEE; IEEE International Conference on Image Processing ; Conference date: 12-09-2010 Through 15-09-2010",

year = "2010",

month = sep,

doi = "10.1109/ICIP.2010.5649331",

language = "English",

}

TY - CONF

T1 - Gradient field descriptor for sketch based retrieval and localization

AU - Hu, Rui

AU - Barnard, Mark

AU - Collomosse, John

N1 - Note: Published in: 2010 IEEE International Conference on Image Processing. Piscataway, NJ : Institute of Electrical and Electronics Engineers. ISSN 1522-4880 ISBN 9781424479924 Organising Body: IEEE

PY - 2010/9

Y1 - 2010/9

N2 - We present an image retrieval system driven by free-hand sketched queries depicting shape. We introduce Gradient Field HoG (GF-HOG) as a depiction invariant image descriptor, encapsulating local spatial structure in the sketch and facilitating efficient codebook based retrieval. We show improved retrieval accuracy over 3 leading descriptors (Self Similarity, SIFT, HoG) across two datasets (Flickr160, ETHZ extended objects), and explain how GF-HOG can be combined with RANSAC to localize sketched objects within relevant images. We also demonstrate a prototype sketch driven photo montage application based on our system.

AB - We present an image retrieval system driven by free-hand sketched queries depicting shape. We introduce Gradient Field HoG (GF-HOG) as a depiction invariant image descriptor, encapsulating local spatial structure in the sketch and facilitating efficient codebook based retrieval. We show improved retrieval accuracy over 3 leading descriptors (Self Similarity, SIFT, HoG) across two datasets (Flickr160, ETHZ extended objects), and explain how GF-HOG can be combined with RANSAC to localize sketched objects within relevant images. We also demonstrate a prototype sketch driven photo montage application based on our system.

KW - Computer science and informatics

U2 - 10.1109/ICIP.2010.5649331

DO - 10.1109/ICIP.2010.5649331

M3 - Paper

T2 - IEEE International Conference on Image Processing

Y2 - 12 September 2010 through 15 September 2010

ER -

Gradient field descriptor for sketch based retrieval and localization

Abstract

Conference

Bibliographical note

Keywords

Access to Document

Fingerprint

Research output

Gradient field descriptor for sketch based retrieval and localization

Cite this