Telenor | Telenor R&D
UiT | Dept. of Computer Science
UiB | Dept. of Information Science and Media Studies
NTNU | Dept. of Computer and Information Science

tromsotromso_pic2tromso2

CAIM





prototypes

CAIM PROTOTYPES

UiT:

  • InfoAlbum is an image centric information album, where the goal is to automatically provide the user with information about i) the object or event depicted in an image, and ii) the surrounding where the image was taken.
    The system aims at improving the image viewing experience by presenting supplementary information such as location names, tags, temperature at image capture time, placement on map, geographically nearby images, Wikipedia articles and web pages.
    The information is automatically collected from various sources on the Internet based on the image metadata gps coordinates, date/time of image capture and a category keyword provided by the user.
    The prototype is documented in “Image Centric Information Collection” (with link)

  • LoCaTagr is a location, category and time-based automatic image tagging system using Flickr. LoCaTagr is able to find relevant tags for a query image as long at there are a sufficient number of geo-referenced and already tagged images available on Flickr that is relevant for the query image. The query image must be geo-referenced, have date/time of image capture available, and the user must provide an image category.

  • ImSE is an image search engine, which automatically annotates images and supports image search in Fronter VLE (Virtual Learning Environment) based on these annotations. The system analyzes the context images are referred in and uses it for automatic image annotation. If an image is used within different contexts, such as different textual documents, it is annotated with the keywords gathered from all contexts. Users search for desired images by typing query keywords in the search field of the systems interface.

  • CbImage Retrieval is a collection-based image retrieval system that infers image semantics for individual images based on information about the collections to which the images belong. Information from these different collections will be combined to create a better description of individual images than is available from the collections they belong to seen separately.
    The system is able to search for images, and is able to combine collection information in order to provide an enhanced view of image semantics for the search process.

  • CollSumm is a system that automatically generates a description of an image collection based on available image metadata. Image metadata within a collection is processed with the goal of locating and extracting representative terms. Representatives are terms that describe the image collection as a whole from its most informative view and that should stand out from other terms in the collection

     

UiB:

  • The MMIR4 - Mobile Multimedia Image Retrieval - prototype extends MMIR3, with 2 new features:
    1) Support for storing the results of a search on the phone so that they can be viewed at a later time and
    2) Selection of a search image from the phone's memory.
    Result sets now consist of pages of 6 images, which better fit the N5800 phone display screen. MMIR4 now returns both a text and audio description of the principle object(s) in the selected image(s). MMIR4 has been developed for Nokia's N5800 m-phone which has a sketch pad interface, and gps. The prototype is documented in Parmann,E. MMIR4 - Mobile Multimedia Image Retrieval (ver.4). 2010.

    •
  • The MMIR3 - Mobile Multimedia Image Retrieval - prototype extends MMIR2 with support for drawing the seed image and selecting a segment of a photograph for use as a seed image. In addition to presentation of result sets of 4 images, the system returns both a text and audio description of the principle object in the selected image(s). MMIR3 has been developed for Nokia's N5800 m-phone which has a sketch pad interface, and gps. MMIR3 is documented in Hellevang,M. MMIR3 - Mobile Multimedia Image Retrieval (ver.3). 2009.

    •
  • The MMIR2 - Mobile Multimedia Image Retrieval - prototype has been developed for the Nokia N95 mobile phone to utilize its GPS positioning and image capture for image retrieval from the BergenBy image database. The prototype is documented in Hellevang, M. MMIR2 - Mobile Multimedia Image Retrieval (ver.2). 2008 .

  • The BergenBy database currently (Nov.2010) contains nearly 600 images of 50+ buildings, statues, monuments, and street scenes within the Bergen city area. Each object annotation includes both a text and audio description. The images include full object photographs, historic photographs, machine created drawings and images of object details, such as a church door or window. The DB is described in both Langøy's 2008 report BergenBy Database & Metadata Editor and in Møller's 2009 report Bergen By - a Multi-Modal Image Database.

  • The VISI4/BergenBy prototype extends VISI3 functionality with support for context aware feedback. Result sets are now presented in pages of 8 images for which the user can mark relevant images. On submission of the images, keywords are presented to the user for further refinement of the result set.
    VISI4 also includes an improved ranking image for CBIR+TBIR searches. The prototype is documented in Døskeland,Ø. VISI4 - Supporting Relevance Feedback in Context Aware Image Retrieval, 2010.

  • The VISI3/BergenBy prototype is a full rewrite in Java of VISI-2 which extends VISI2 functionality with presentation of more information about the images in the result set. VISI3 also includes an improved ranking image for CBIR+TBIR searches. The prototype is documented in Carlson, C. VISI3 - Context Aware Image Retrieval. 2009

  • The VISI2/BergenBy prototype extends VISI-1 with support for utilization of context variables in visual image retrieval. Currently, VISI2 supports image retrieval using image content (CBIR), text descriptors (TBIR) and GPS location specifications. Aslo supported are CBIR+GPS and CBIR+TBIR input combinations.

  • The VISI-1 prototype provides an image retrieval system for exploring basic CBIR (Content-based image retrieval) techniques as implemented in Oracle's image search functions. It was adapted for the CAIM project in the summer of 2007 by application of the initial Bergen/By database.

  • The VISI/Maritime prototype supports visual image retrieval (CBIR) using uploaded images or user constructed drawings.

  • The Maritime database contains 400+ images of predominently, marine animals.

 

Telenor:
    The Telenor prototypes are described in Sigmund Akselsen, Bente Evjemo and Anders Schürmann. (2008) CAIM Prototypes. CAIM project meeting, Tromsø 25.09.2008 Telenor R&I, Products and Markets
  • M2S - Tourist information in multiple channels(images of info guide ads)
  • TIFF - Tromsø International Film Festival event assistant (barcodes).
  • Smart Binoculars - Points of Interest info in images (camera position, direction and depth of field)
  • Visual search client for iPhone (general content based image recognition)
  • TVG - Tromsø Visual Guide (images of sculptures and buildings, position, ads)

 

Created © University of Tromso