Bag of Words Models for visual categorization,[url=http://www.outletonline-michaelkors@#$%&/]michael kors purse[/url]
Lately, I been reading a lot about BOW (Bag of Words) models [1] and I thought it would be nice to write a short post on the subject. The post is based on the slides from Li Fei Fei taken from ICCV 2005 course about object detection:
As the name implies,[url=http://www.giuseppe-zanotti.us@#$%&/]guiseppe zanotti[/url], the concept of BOW is actually taken from text analysis. The idea is to represent a document as a of important keywords, without ordering of the words (that why it a called of words instead of a list for example).
We can use the bag of words model for object categorization by constructing a large vocabulary of many visual words and represent each image as a histogram of the frequency words that are in the image. The following figure illustrates this idea:
Bag of words representing object as histograms of words occurrences
How exactly do we construct the model? First,[url=http://www.mulberryhandbags-outlet.co.uk/]mulberry bags[/url], we need to build a visual dictionary. We do that by taking a large set of object images and extracting descriptors from them (for example,[url=http://www.nikeair-max.fr/]@#$%& air max thea[/url], SIFT on a grid or from detected keypoints (see the post on descriptors if you not sure about their usage).
Next, we cluster the set of descriptors (using k means for example) to k clusters. The cluster centers act as our dictionary visual words.
Given a new image, we represent it using our model in the following manner: first,[url=http://www.lululemon-outlet.us@#$%&/]www.lululemon@#$%&[/url], extract descriptors from the image on a grid or around detected keypoints. Next,[url=http://www.michaelkorsoutlet-online.us@#$%&/]michael kors backpack[/url], for each descriptor extracted compute its nearest neighbor in the dictionary. Finally, build a histogram of length k where the i value is the frequency of the i dictionary word:
Bag of words representing object as histograms of words occurrences
This model can be used in conjunction with Nave Bayes classifier or with an SVM for object classification[1],[url=http://www.polo-ralphlauren.it/]polo @#$%&[/url].
There is a nice demonstration in Vlfeat of a SIFT based BOW model and SVM for object classification on the Caltech101 benchmark.
I ran a tiny example of the code using only 10 classes,[url=http://www.oakleyvaultsunglasses.us@#$%&/]oakley gascan[/url], 15 images for training and 15 images for testing and got the following confusion matrix:
The confusion matrix shows the score that each classes training set got when running each of the classes classifiers on them the (i,j) element is the result of applying class j classifier on class i training set. As you can see,[url=http://www.rolex-replicawatches.us@#$%&/]vintage rolex[/url], the top scores are on the diagonal,[url=http://www.louboutinoutlet.it/]christian louboutin[/url], meaning the each class top score was obtained where running each class classifier on its training data.

[url=http://www.shejianfarm@#$%&/bbs/forum.php?mod=viewthread&tid=308125&extra=]@#$%& rosh run 3 Proven Tips To Look Like A Winner[/url]

[url=http://www.aps.sn/newsedit/spip.php?article43524]babyliss pro nano titanium Aide Apologized For Child Porn[/url]

[url=http://www.scdesignconcept@#$%&/_mickaelphp5/picture.php?/669/list/769,713,669,727,791,772,677,509,749,659,685,283,656,68,787/start-2430&comments_order=DESC#comments]vanessa bruno sac AirTran did the best job for pas[/url]

[url=http://www.boffj@#$%&/viewthread.php?tid=262711&extra=]longchamps Armstrong bags Galway a share of spoils[/url]

[url=http://www.bluegrasscollege.org:8080/jforum/posts/list/0/30695.page#30935]sacs longchamp Best airlines for bringing sports gear[/url]