[
http://jira.nuxeo.org/browse/NXSEM-9?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Olivier Grisel updated NXSEM-9:
-------------------------------
Description:
Same feature as NXSEM-8 but for picture content. Instead of using TF-IDF
features we can use normalized gray level pixel values of a scaled down
(100x100 or 500x500) version of the picture pre-propecessed using the GIST
global picture descriptors [1] using the LEAR C implementation (GPL) and maybe
post-ranking using leveraging a smarter features extraction layer based on SIFT
[3] using the libsiftfast [4] library.
[5] provides a nice state of the art of related algorithms with scalability
numbers.
The goal is to be able to lookup related pictures in DAM (right click + "find
more pictures like this one") or to perform picture query by example.
[1] http://people.csail.mit.edu/torralba/code/spatialenvelope/
[2] http://lear.inrialpes.fr/~jegou/src.php
[3]
http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/AV0405/MURRAY/SIFT.html
[4] http://sourceforge.net/projects/libsift/
[5] http://lear.inrialpes.fr/pubs/2009/DJSAS09/
was:
Same feature as NXSEM-8 but for picture content. Instead of using TF-IDF
features we can use normalized gray level pixel values of a scaled down
(300x300 or 500x500) version of the picture or leveraging a smarter features
extraction layer based on SIFT [1] using the libsiftfast [2] library.
The goal is to be able to lookup related pictures in DAM (right click + "find
more pictures like this one") or to perform picture query by example.
[1]
http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/AV0405/MURRAY/SIFT.html
[2] http://sourceforge.net/projects/libsift/
> CoreEventListener + service to build 64bits semantic hash of documents with
> picture content
> -------------------------------------------------------------------------------------------
>
> Key: NXSEM-9
> URL: http://jira.nuxeo.org/browse/NXSEM-9
> Project: Nuxeo Semantic R&D
> Issue Type: Task
> Reporter: Olivier Grisel
> Assignee: Olivier Grisel
>
> Same feature as NXSEM-8 but for picture content. Instead of using TF-IDF
> features we can use normalized gray level pixel values of a scaled down
> (100x100 or 500x500) version of the picture pre-propecessed using the GIST
> global picture descriptors [1] using the LEAR C implementation (GPL) and
> maybe post-ranking using leveraging a smarter features extraction layer based
> on SIFT [3] using the libsiftfast [4] library.
> [5] provides a nice state of the art of related algorithms with scalability
> numbers.
> The goal is to be able to lookup related pictures in DAM (right click + "find
> more pictures like this one") or to perform picture query by example.
> [1] http://people.csail.mit.edu/torralba/code/spatialenvelope/
> [2] http://lear.inrialpes.fr/~jegou/src.php
> [3]
> http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/AV0405/MURRAY/SIFT.html
> [4] http://sourceforge.net/projects/libsift/
> [5] http://lear.inrialpes.fr/pubs/2009/DJSAS09/
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://jira.nuxeo.org/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
_______________________________________________
ECM-tickets mailing list
[email protected]
http://lists.nuxeo.com/mailman/listinfo/ecm-tickets
