OntoGen Extension for Exploring Image Collections
0 likes327 views
The document discusses the challenges of handling high-dimensional image data and presents Ontogen as a tool for semi-automatic ontology construction and image analysis through feature extraction and clustering techniques like k-means. It outlines the process of extracting features, such as SIFT and color histograms, from images and how these can be used to organize and refine image collections. The results indicate that a semi-supervised approach can effectively construct image ontologies by detecting and refining concepts based on their visual features.
OntoGen Extension for Exploring Image Collections
- 1. VISUALIZING IMAGE COLLECTIONS WITH ONTOGEN From Images To Ontologies
- 2. IMAGE DATA Difficult to handle High-dimensional representations The amount of image data is constantly increasing and there is a rising need for reliable automatic image analysis systems in practical applications
- 3. Image representation Application Data Mining Extract features Text Color info SIFT features
- 4. SIFT FEATURES Rotation, scale and translation invariant orientation gradients located at “interesting” points on an image Usually, the SIFT feature space is quantized so that some “representative” vectors are found Each feature on an observed image is then assigned to its nearest representative and this is how the so called “codebook” histogram is obtained
- 5. COLOR HISTOGRAMS Color information on an image might or might not be of interest for a particular problem, but it usually represents a useful piece of information There are several ways to handle this information, but the simplest and fastest one is to simply divide the color spectrum into “buckets” and calculate the distribution of colors into these buckets, thereby obtaining the color histogram for an image
- 6. ONTOGEN OntoGen is a tool which allows us to do semi- automatic ontology construction, clustering, classification, as well as data visualization via multidimensional scaling This can easily be applied on image data to gain an overview of collections of images
- 7. IMAGE FEATURE EXTRACTION We extract SIFT features and color histograms for each image We calculate the distance between images as the weighted sum of distances between the two distributions (SIFT codebook and color data) If images have annotations, this can easily be incorporated by adding a third part in the representation for each image
- 8. ONTOGEN ON IMAGE DATA On the next few slides we show the usage of OntoGen on one simple data set The data was taken from ImageNet online image collection. The particular subset contains images of various types of flowers, as well as images of fire and images of buildings
- 9. MAIN WINDOW WHEN THE COLLECTION IS LOADED
- 10. DOCUMENT LIST FOR QUICK OVERVIEW
- 11. DOCUMENT ATLAS WHEN NOT DISPLAYING IMAGES
- 12. DOCUMENT ATLAS WHEN DISPLAYING IMAGES
- 13. CREATING AN ONTOLOGY We can do k-means clustering to detect groups of similar images We can use these groups to create a level in the ontology The relevant features are displayed on top of the nodes
- 14. SO, LET’S LOOK AT SOME OF THOSE NODES AND THEIR MEDOIDS…PRETTY GOOD…
- 15. HOWEVER… One of the first-level sub-concepts is not good, which can be seen by observing it’s medoids: So, now we can branch it further into more refined sub-concepts to improve the quality
- 16. BEFORE WE DO SO, WE CAN VISUALIZE THE SUB-CONCEPT IN DOCUMENT ATLAS
- 17. SO … This is definite evidence that the concept should be split into at least two different sub-concepts Most of the images inside it represent buildings, but there are some that belong to a certain type of flower, as well as some depicting fire So, just to be safe, let’s say we want 5 sub- concepts
- 18. THIS IS HOW THE NEW ONTOLOGY WILL LOOK LIKE:
- 19. AND THE MEDOIDS FOR THE FIVE NEW REFINED SUB-CONCEPTS ARE:
- 20. CONCLUSIONS What we see is that we can construct an image ontology in a semi-supervised way By using k-means clustering based on SIFT+color image representation we can detect candidates for concepts in the ontology and then refine them until we reach good quality
- 21. AKNOWLEDGEMENTS Thiswork was supported by the bilateral project between Slovenia and Romania “Understanding Human Behavior for Video Survailance Applications,” the Slovenian Research Agency and the ICT Programme of the EC PlanetData (ICTNoE-257641).