Thread Rating:
  • 1 Votes - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Bootstrapping visual categorisation with negative relevance (JAVA) .
07-04-2014, 09:45 PM
#
Bootstrapping visual categorisation with negative relevance (JAVA) .
Bootstrapping Visual Categorization
With Relevant Negatives

ABSTRACT

Learning classifiers for many visual concepts are important for image categorization and retrieval. As a classifier tends to misclassify negative examples which are visually similar to positive ones, inclusion of such misclassified and thus relevant negatives should be stressed during learning. User-tagged images are abundant online, but which images are the relevant negatives remains unclear. Sampling negatives at random is the de facto standard in the literature. In this paper, we go beyond random sampling by proposing Negative Bootstrap. Given a visual concept and a few positive examples, the new algorithm iteratively finds relevant negatives. Per iteration, we learn from a small proportion of many user-tagged images, yielding an ensemble of meta classifiers. For efficient classification, we introduce Model Compression such that the classification time is independent of the ensemble size. Compared with the state of the art, we obtain relative gains of 14% and 18% on two present-day benchmarks in terms of mean average precision. For concept search in one million images, model compression reduces the search time from over 20 h to approximately 6 min. The effectiveness and efficiency, without the need of manually labeling any negatives, make negative bootstrap appealing for learning better visual concept classifiers.





Existing System
Traditionally, labeled examples are annotated by expert annotators.However, expert labeling is labor intensive and time consuming, making well-labeled examples expensive to obtain and consequently their availability is limited.

Disadvantage:
Labor intensive and expensive.
Time consuming.

Proposed System:
Much research has been conducted towards inexpensive solutions to acquire positive examples, e.g., from web image search results or socially tagged data, or by online collaborative annotation. For instance, Schroff train a visual classifier on web image search results of a given concept, and re-rank the search results by the classifier. Though the automated approaches are not comparable to dedicated manual annotation, their output provides a good starting point for manual labeling. Deng et al build an ImageNet wherein positive examples of a WordNet concept are obtained by labeling web image search results of the concept using a micro payment service. Compared to traditional expert labeling, the new labeling mechanism yields positive examples for many categories with lower cost. In this paper we assume that positive examples are obtained by (one of) the approaches described above, and focus on obtaining negative examples.
Advantages:
Lower cost.
Inexpensive.
.



Problem Statement
Since negative examples of a concept belong to many other concepts, most of expert labeling efforts are dedicated to annotating negatives. One might consider bypassing the negative labeling problem by one-class learning, which creates classifiers using positive examples only. However, because negative examples also bear valuable information, they are important for classification. This has been well observed in Tao et al for interactive image retrieval. Our empirical study shows that visual classifiers trained by one-class learning are inferior to classifiers trained by two-class learning. So labeling negatives remains essential, but methods which can reduce the manual labeling effort are needed.


Scope:
For the implementation of asymmetric bagging, we follow but use the full feature space rather than random subspaces, as studying random subspaces is beyond the scope of this paper. In the -th iteration, random sampling selects at random to train a classifier, while asymmetric bagging uniformly combines the classifier and t-1 classifiers generated in the previous rounds.
To reduce the chance of incorrectly selecting genuine positives for a given concept , we remove images labeled with or its semantically related tags. We observe that if an image is labeled with visual concepts, but not labeled with or its semantically related tags, the image tends to be a negative example of w.








Architecture:



MODULES”
1. Positive Examples.
2. Support Vector Machine.
3. Asymmetric Bagging.
4. Negative Bootstrap.

Modules Description
1. Positive Examples

When the number of concepts is large, obtaining labeled examples in an efficient way is essential. Here, labeled examples are annotated by expert annotators. However, expert labeling is labor intensive and time consuming, making well-labeled examples expensive to obtain and consequently their availability is limited. Much research has been conducted towards inexpensive solutions to acquire positive examples from web image search results or socially tagged data, or by online collaborative annotation.

2. Support Vector Machine

In machine learning, support vector machines are supervised learning models with associated learning algorithms that analyze data and recognize patterns, used for classification and regression analysis. The basic SVM takes a set of input data and predicts, for each given input, which of two possible classes forms the output, making it a non-probabilistic binary linear classifier. Given a set of training examples, each marked as belonging to one of two categories, an SVM training algorithm builds a model that assigns new examples into one category or the other. An SVM model is a representation of the examples as points in space, mapped so that the examples of the separate categories are divided by a clear gap that is as wide as possible. New examples are then mapped into that same space and predicted to belong to a category based on which side of the gap they fall on.








3. Asymmetric Bagging

The robustness of the final classifier is improved by classifier aggregation, the quality of asymmetric bagging is bounded by the lack of relevant negatives. Moreover, since all meta classifiers need to be executed, the classification time is proportional to the amount of meta classifiers. In sum, the lack of relevant negative examples and the computational cost for running all meta classifiers put the effectiveness and efficiency of asymmetric bagging into question.



4. Negative Bootstrap

Given a set of unlabeled images, we search for images which contain a specific concept by employing a visual classifier of the concept. Let be an image. Its content-based representation
is a -dimensional feature vector. We will refer to an image and its corresponding feature vector interchangeably, using x(i) to indicate the i-th dimension of the vector. We use g(x) to denote a classifier, which produces a real-valued score of an image being a positive instance of the target concept. In particular, g(x) >0 means the image is classified as positive, and negative otherwise.







System Configuration:-
H/W System Configuration:-

Processor - Pentium –III

Speed - 1.1 Ghz
RAM - 256 MB (min)
Hard Disk - 20 GB
Floppy Drive - 1.44 MB
Key Board - Standard Windows Keyboard
Mouse - Two or Three Button Mouse
Monitor - SVGA
S/W System Configuration:-

 Operating System :Windows95/98/2000/XP
 Application Server : Tomcat5.0/6.X
 Front End : HTML5,CSS3, Java, Jsp
 Scripts : JavaScript.
 Server side Script : Java Server Pages.
 Database : Mysql
 Database Connectivity : JDBC.


Help me to do this ......


Attached File(s)
.pdf  Bootstrapping Visual Categorization With Relevant Negatives.pdf (Size: 2.58 MB / Downloads: 29)
 


Possibly Related Threads...
Thread: Author Replies: Views: Last Post
  JAVA information system gopal gautam 0 11 Yesterday
Last Post: gopal gautam
  TCP/IP MONITORING AGENT USING JAVA yehmiesneh 0 30 11-19-2017
Last Post: yehmiesneh
  MakeUP_MS Java/SQL emnolina 0 118 05-21-2017
Last Post: emnolina
  Fast nearest neighbour search project in dotnet or java mukund 0 178 04-09-2017
Last Post: mukund
  Online art gallery in java Bhagyashree Vedpathak 0 319 03-25-2017
Last Post: Bhagyashree Vedpathak
  java projet working in eclips Hina Vaishnavi 0 157 03-11-2017
Last Post: Hina Vaishnavi
  online monthly stock management system in java and jsp Prince Ahuja 0 174 03-05-2017
Last Post: Prince Ahuja
  Photo Management System in JAVA Nicolas JQ 0 542 09-10-2015
Last Post: Nicolas JQ
  Email Application in java Rupali Bhalla 0 544 06-19-2015
Last Post: Rupali Bhalla
  fingerprint employee attendace and time android source code with java getawdejen 0 517 11-22-2014
Last Post: getawdejen

Forum Jump:


User(s) browsing this thread:
1 Guest(s)

Return to TopReturn to Content