Visual information retrieval using synthesized imagery Thomee 2007 Leiden
<p>Jan 29, 2011 11:48 AM
 
In this project (VIRSI) we investigate the promising <b>content based</b>
<b>retrieval paradigm known as interactive search or relevance</b>
<b>feedback</b>, and aim to extend it through the use of s<b>ynthetic</b>
<b>imagery. </b>In relevance feedback methods, the user himself is a key
factor in the search process as he provides positive and negative
feedback on the results, which the system uses to iteratively
improve the set of candidate results. In our approach we closely
integrate the generation of synthetic imagery in the relevance
feedback process through a new fundamental paradigm: <b>Artificial</b>
<b>Imagination (AIm</b>).
 
 Content-based retrieval, PCA= prin comp analys,  artif imagination
 
...  may be necessary to directly analyze the PICTORIAL CONTENT of the media.
 
Our own visual imagination allows us to create new and useful
examples based on our memories and experiences. In artificial
imagination we intend to harness similar strategies to intelligently
synthesize informative examples. In particular, we intend to
endow the computer with the ability to ask whether a particular
synthesized example (which is not yet in the database) is relevant
 
One example from real life is
when a <b>police officer creates a sketch of an unknown person to</b>
<b>help identify a criminal</b>.
 
HX of systems  QBIC  and Virage  mid 90s.
 
The resulting relevance feedback
methods  usually share the
following process: <b>the system shows the user a set of candidate
results; the user labels a subset of the candidates as positive or
negative examples</b>; based on these examples the system
<b>reformulates the representation of the user query,</b> and
subsequently<b> presents the user with the next set of candidate
results.</b> This process is repeated until, hopefully, the user is
satisfied. Relevance feedback can be considered a special case of
<b>emergent semantics.</b> Other names used in the computer vision
literature have included <b>query refinement, interactive search, and
active learnin</b>g.
 
A general approach is to view<b> relevance feedback as a type of
pattern classification where a relevant class is learned from a set
of training examples with relevant and irrelevant labels. In
principle, it is therefore possible to apply any type of learning
algorithm in the relevance feedback loop</b>.
 
small training setss... a problem
used AdaBoost and SVM... some combo.  SVM = support vector machine
 
In our system...  we focus on the feedback-based generation of examples:
we use the classic and wel-known relevance feedback method proposed by Rocchio
the simple idea is to <b>move a query point toward the relevant examples</b> and 
<b>away from the irrelevant</b> example.
 
section 3:  ARTIFICIAL IMAGINATION
 
<b>Our visual imagination allows us to create newly synthesized</b>
<b>examples based on our memories and experiences</b>. When we are
<b>learning new visual concepts, we often construct such examples</b>
<b>based on real objects or scenes to help understand or clarify the</b>
<b>primary features which are associated with the concept</b>. As
mentioned, in this project we aim to explore the way in which
<b>artificial imagination by intelligently computed examples can be</b>
<b>used to aid a computer algorithm in learning new visual concepts</b>,
and in particular to obtain meaningful solutions to user queries on
multimedia databases.
 
examples are synthesized by the system to satisfy info requirements.
 
We thus understand “artificial imagination" here as the intelligent
synthesis of examples based on known database entries and their
location in the feature space.
 
See example figures:  the figures include  synthetic images (not in the database)
to guide the user toward his goal.
 
The main future challenge lies in the design of meaningful methods for image synthesis.
 
 
.
 
 
 
 
examples.
 
 
 
 
 
 
 </p>