[prev in list] [next in list] [prev in thread] [next in thread] 

List:       monetdb-users
Subject:    [MonetDB-users] Locality Sensitive Hashing method for Image Retrieval
From:       BOOM Bas <bboom () inf ! ed ! ac ! uk>
Date:       2012-06-18 9:28:48
Message-ID: 4FDEF4D0.2070808 () inf ! ed ! ac ! uk
[Download RAW message or body]

Hi,

Sorry, I am new to this mailinglist and not sure this question fits, but 
have I question how to optimize my queries

I am experimenting with monetdb for image retrieval, having an image 
database of 300000 image.
At the moment, I am using the Locality Sensitive Hashing method and 
because I understood from the
monetdb documenation that monetdb supports indexing/hash tables, which 
makes the implementation
quiet easy in monetdb.

I have for each image 30 buckets with 24 bit strings (int) (should be 
indexed, assume automatically done by monetdb)
I have for each image a feature vector of 1000 real values.
I have for each image a unique id together with some info like file name.

At the moment, my table is as follows

image_id, filename, bucket_1, bucket_2, bucket_3 ... bucket_30, 
feature_value_1, feature_value_2, ..., feature_value_1000

For Locality Sensitive Hashing, I am performing the following sql query:

select * from table where bucket_1 = bitstring or bucket_2 = bitstring 
or bucket_3 = bitstring or ... bucket_30 = bitstring (limit 1000)

These bitstrings are quiet unique, giving me around the 500 images from 
the entire database.
I would also like to obtain the feature vector (so all feature values). 
At the moment, query and data retrieval times depends but often take 
around 0.5-1 minute, what are ways to improve this. Does using 2 tables 
help or storing the feature vector in one column? Ideas are welcome.

I am using the MAPI client to retrieve the data. Happy to provide more 
information if necessary

Cheers,

Bas Boom
University of Edinburgh
part of Fish4Knowledge consortium




-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
MonetDB-users mailing list
MonetDB-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/monetdb-users
[prev in list] [next in list] [prev in thread] [next in thread]