K-Space kick-off meeting

11-13 January 2006, London (UK)

Author: Lynda, Raphael
CWI participants: Lynda, Raphael
# participants: around 40

Table of Contents

1. Partner introductions:

  1. QMUL, UK (coordinator): Multimedia and Vision Lab
    2. EU IP: involved in aceMedia, and is a core partner of MESH - Multimedia Semantic Syndication for Enhanced News Services (start, March 2006)
  2. Koblenz, DE
    3. EU projects: involved in NeOn (ontologies), X-Media, aceMedia (Multimedia), and web services or P2P related projects
  3. Joanneum Research, AT: Digital Media group
    4. Non-profit research company located in Gratz. Applied research: bridge between university and market (sort of TI) 75% funded by industry. Work on AV content, and develop MPEG-7 tools (their competencies).
  4. ITI-CERTH, GR: Multimedia Knowledge Group + NTUA, GR: Image, Video and Multimedia Systems Lab (IVML)
    5. EU projects: involved in aceMedia and KnowledgeWeb + the forthcoming projects MESH, X-Media and Boemie (all start March 2006)
  5. DCU, IR:
    6. Co-ordinator of TREC Video. EU projects: involved in aceMedia, MultiMatch (Multimedia and multilingual access to cultural heritage) and Cost 292
  6. CWI, NL:
  7. GET-ENST, FR: Signal, image and sound processing group
  8. INA, FR: Audio-visual Content Description Group
  9. Eurecom, FR: Mobile Communications and Multimedia Communications Groups
  10. GU, UK: Information Retrieval Group
  11. DFKI, DE:
    12. EU projects: involved in MESH. Competencies: language technologies and Semantic Web, + a little bit of MM indexing
  12. TUB, DE: Communication System Group
    13. Strong participation in Content analysis, Audio speech analysis
  13. EPFL, CH
    14. Would like to not emphasize MPEG-7, and plan to send reports of MPEG activities
  14. UEP, CZ: Department of Information & Knowledge Engineering (DIKE)
    15. Main research areas: NLP, information retrieval, soft computing ... and now Semantic Web

2. WP1 - Administration, Management (QMUL)

2 mailling lists (kspace + management board)

Template Approval: Template naming system: ks-WP012-01-[loc]-[date]-[verbose].doc Public Web Site (public): Project Management Site (internal) Other IST Projects: Forms:

3. WP2 - Integration Activities (DFKI)

Dissemination: invite other non-K-Space people out of own budget and declare it on own K-Space money. To get the extra WP2 exchange money you need to send a PhD student to somewhere. 6 short term fellowships. 15 studentships by month 18. Always be thinking of it - need one per partner. Set up mentality. Discuss about WP 2.3 on how to organise the web site. (Presentation issues and meta-data, also search from Alan's point of view.)

4. WP7 - Activities to Spread Excellence (JRS)

5. WP8 - Assessment and Evaluation (Eurecom)

Annual project review: All partners, input from other WPs, and will establish the program of work for year 2.

6. WP6 - Framework for the integration of software tools (DCU)

Need to facilitate research resource sharing: easy exchange of tools, interfaces, test data and results. One of the objective is to put in place the software and network interfaces to ensure that partners have access to and can use the research resources of the other partners.

The output of this WP can be thought as the physical instanciation of the K-Space.

Strategy: a NoE is not an IP or a Strep:

7. WP3 - Content-based multimedia analysis (JRS)

Strong links with WP4 and WP5 ! There was a discussion of each sub task. For each sub-tasks, all the partners involved detailed their background, expertise, technology and proposed some research challenge. The key point is to see how several partners can work together on particular issue. The key word is "collaboration".

Each task should more or less produce 2 differents deliverable in the first 18 months: a state of the art report and either a research plan with proposed new approaches and/or a set of tools

The CWI do not participate to this WP

  1. WP 3.1: Content Structuring (Benoit Huet, Eurecom)

    2. The goal is to reveal the structure of content items to help in inferring semantic information. The "Ultimate goal" is to decode this "grammar of moving images"

  2. WP 3.2: Moving 2D and 3D Object Segmentation and Indexing (NoelO'Connor, DU)

    3. Categorization and description of objects, and describe their interactions

  3. WP 3.3: Audio/Speech Processing and Text Analysis (Gaël Richard, GET)

    4. Audio semantics analysis, developped low-level ontology, method for classification, apply textual analysis on OCR.

    Goals: Provides:
  4. WP 3.4: Content Description (Peter Schallauer, JRS))

    5. Develop tools to describe all metadata of WP3 in MPEG-7, extend existing descriptors and description schemes

    Goals: Workplan:

Personal impression: many partners discover here the (too) strong role of MPEG-7 and begin to be worried. The risk is the MPEG-7 world (signal analysis) and the SW world (knowledge representation) ignore themselves ... Each WP (3, 4, and 5) will have this interoperability task => but this should be just one activity (my proposal !)

Interesting discussion: low-level features will be encoded in MPEG-7, other KR formalism will encode higher semantics and all these different should remain interoperable (work in charge of the WP3, 4 and 5).

8. WP4 - Knowledge extraction (ITI)

The goal is to bridge the semantic gap between extracted low-level features and higher semantics concepts using a multimedia ontology infrastructure.Here, Semantic Web technologies + rules should be used for knowledge representation.

CWI will participate to Task 1, 2, 4 and 5

  1. WP 4.1: Specification of a Multimedia Ontology Infrastructure

    2. Should develop a top-level ontology, a low-level ontology, and a knowledge base prototype

    CWI contribution: Expertise on building multimedia ontologies / link to MM TF

  2. WP 4.2: Knowledge-assisted multimedia analysis

    3. Transition from low-level features to symbolic. Deals with the semantic object detection, exploiting the ontologies of WP 4.1, and the content processing algorithms of WP3.

    CWI contribution: weak, not our research activities, we can just provide some hints as for the understand/use of the ontology

  3. WP 4.3: Knowledge-assisted Multimedia Reasoning and Annotation

    4. Automatic semantic annotation of the content: Deals with the extraction of meaningful interpretation of high level events and automatic semantic annotation of multimedia content.

    CWI contribution: We do NOT participate to this sub-task

  4. WP 4.4: Context Based Multimedia Mining

    5. Gather contextual data from the user interaction, group the images in collections, use this grouping for improving the search, etc ... => Obscur goal, even the corrdinator do not understand what this task deals with!

    I argue that this is really content-based mining, means deals with the content and not the end user for providing browsing, summarization of raw collections

  5. WP 4.5: Intelligent Image Classification and User Relevance Feedback

    6. Goal: derive semantic information from past-experience using relevance-feedback. Binary classification problems using SVM + biologically inspired classification problem.

    CWI contribution: links with WP 5.3, expertise as for exploiting human-machine interaction (talk about the work Lloyd do ...)

    Other partners contributions:

9. WP5 - Semantic Multimedia (KU)

The Main objective: how to represent (Task 1) multimedia content and context, such that the further distribution (Task 2) analysis (Task 4) and interaction (Task 3) with content and context to help understand/process multimedia

CWI will participate to Task 1 and 2. CWI will lead the Task 3

CWI should determine which kind of expressive power we need for querying => will influence the need for reasoning, rules, etc ... Since It seems a hard task and we should clarify that !

Organisational issues:
  1. WP 5.1: Knowledge Representation for Multimedia

    2. Extension of RDF/OWL/SWRL: Each partner should write his background in terms of expressivity need for representing multimedia metadata

    Propose to:

    CWI has 8 months on this task

  2. WP 5.2: Reference Framework for Distributed Semantic Management of Multimedia Metadata

    3. Store content and context in a metadata repository, implement a centralized repository, and specify a distributed version. The vision is to challenge Flickr !

    Proposed approach: Ambiguity: what kind of information should be stored here ? Only semantic annotation ? Or the MPEG-7 low level metadata ?

    CWI has 8 months on this task

  3. WP 5.3: Semantic-based Interaction with Multimedia

    4. Semantic retrieval (query and browse) and semantic-based presentation of multimedia (adapted to the user)

    Nobody has a clear idea of what to do here, so CWI (leader) can develop his ideas and close relationships with the work done in MultimediaN

    CWI has 18 months on this task

  4. WP 5.4: Knowledge Extraction from Complementary Sources

    5. Augment the semantic multimedia metadata basis, using primary, secondary and tertiary sources attached to the media.Analyse complementary textual resources (crawled on the web, close caption, teletext, EXIF for pictures, etc ...)

    CWI has 0 month on this task

10. Technical Management Committee Meeting

Clustering with other projects: aceMedia (DCU, ITI, one other) ; 3D TV ; AIM@SHAPE Tomas Sikora, 3D ; X-Media ; MMKM

Video content, K-Space signs for data. Then individuals can ask for data from K-Space.

EPFL: two standards. MPEG-7, JPG ! Never took off. Why? Lots of non-technical reasons. MPEG-7 standardises too little of a system. JPG decided to create JPsearch (not C search). Metadata of MPEG-7 and adding Another standard CB search (content based search) and multi-modal

Scientific advisory board:

If there are problems with entering the (bi-)monthly report information then we can have the tool improved. (For a while.) George should be able to enter all the details.

11. Final Notes

A copy of the WP presentations should be available but it is still not the case ...

11.1 Project Officer meeting

Albert Gauthier (EU): albert.gauthier@cec.eu.int

Amendments: All must be submitted by letter. First email with draft letter then final letter. All amendments accepted or rejected in one go.

If we publish a paper representing K-Space then we need to ask permission (through Craig) first. If we are just submitting a paper acknowledge K-Space funding at the end then we just get on and submit and publish it. 30 days notice prior to publication.

FP6 is lot more freedom with money (cf FP5). It is up to us - don't ask him!

Need to become visible - central point for our objectives. Conference. Papers (in name of project) also important. To external world and to commission. Send published papers to commission.

NoE Showcase, demonstration. No tool demonstration. Deliverables - align with technical annex. Fulfill, integration reseachers etc. AIM@SHAPE has a leading website with lots of shapes. Other project no software. Difficult to assess work done by NoE. Reviewer has difficult task to say yes or no.

IST conference: Helsinki, November. Will NoE get an invitation. Cordis website. Albert in charge of one session for the unit. He will need people to talk. IBC session Our conference Nov/Dec - use this conference to publicise FP7 One day symposium day before or after conference. Has to be easily reachable by people.

FP7: Added partners outside Europe, e.g. USA.. If they add value and are important and this is clear in annex, then that is fine. (Used to be the small Greek partner to "balance" the 6 UK and German partners. Don't use the US partner just for making proposal more exotic.)

11.2 Political considerations

Raphael pessimist view of the meeting: