It was concluded that the applications document needs cleaning up. In the absence of Klaus Diepold (and the uncertain return) the Requirements Group was very pleased to receive an offer from Fernando Pereira to update the document after the meeting.
Version 2 related Requirements issues
Requirements for fine-grain scalability were confirmed and slightly extended in the MPEG 4 Requirements document. The primary goal is streaming over links for which the bitrate is not known in advance, such as over IP networks. It was noted that such ‘fine grain scalability’:
should work with pre-encoded material;
should work in multicast environments.
Based on core experiments, it will be seen whether there is technology that can satisfy these requirements (which will of course be after Version 1).
Because there was discussion about the practical applicability of the fine grain scalability, a resolution was adopted as follows:
The Requirements Group recommends that a study be made to the applications and operational environments for fine-grain scalability, to help deciding which tools suit the requirements. In particular, this applies to how the proposed tools would work together with transport protocols used on the Internet.
Requirements on 3D meshes were updated in the Requirements Document. It was confirmed that progressive download and error resilience -general MPEG-4 objectives are also important for 3D meshes in particular. The Requirements Group was pleased to learn that the technology is likely available in the SNHC group.
A new MPEG-4 Overview was drafted quickly after the meeting, based on contributions and remarks received during the meeting, and on the changes to the standard introduced during the meeting. There is now only one single Overview, which covers both Versions 1 and 2.
MPEG-7 Requirements issues
Again the issue of push versus pull was discussed. The conclusion was that there was no impact on the Requirements Document. Also again, the issue of real-time vs. non real time was brought to the discussion. Again it was decided that there is no need for a split in two tracks. The Ad Hoc Group on MPEG-7 Requirements would be the place to discuss whether requirements were adequately addressed for real-time database applications.
The definition of a DS was changed to allow hierarchical nesting of DSs.
Examples were included in the Main body rather than in annexes.
The issue of embedding code was discussed again. The conclusion was that it may or may not be needed, but that the requirements for it are unclear, and that nothing can be done until they are clear. Also, it was noted that a similar thing was developed in MPEG-4, but that this was done at a later stage. Not all problems need to be solved at the same time.
IPR was confirmed to be an issue for MPEG-7 Descriptions too, and of course MPEG-7 descriptors need to be able to identify IPR on content itself. (Note that MPEG-4 OCI provides a solution for this, that was much debated with the creative industries).
Storing user preferences was recognized as an important issue in MPEG-7 enabled applications, but not as a prime responsibility for MPEG-7 at the moment.
New requirements for the DDL were identified. Among them are the capability to describe spatial and temporal links, and inheritance.
Common (Intermediate) Language?
It was clarified that textual descriptions should support all languages (i.e. international character sets). Note that MPEG-7 will not focus on descriptors for text. It was noted that ‘a textual descriptor’ (for possibly non-textual content) is not the same as ‘a descriptor for text’.
The desirability of a ‘Common Intermediate Language’ was discussed. Although everybody was sympathetic with the idea, people with experience in the field (and there were quite a few around) explained why this is very hard - or even impossible - to achieve.
In the light of this discussion, a requirement was added that MPEG-7 should support textual descriptions in multiple languages, and that it should be possible to indicate that the descriptions are supposed to be multiple representations of the same description.
In a discussion and a joint meeting with Systems, the possible need for an MPEG-7 Systems layer was discussed. Olivier Avaro explained that many features could be provided by MPEG-4 Systems/DMIF, among which:
synchronization and multiplexing of different MPEG-7 and content streams (implementation, not just a specification like SMIL)
abstraction from service model (broadcast or on-line)
Note that MPEG-4 Systems has already defined a streamtype called ‘MPEG-7’.
The conclusion of the discussion was that the AHG on MPEG-7 Requirements would have a study item ‘Requirements for MPEG-7 Systems’.
The duplications between the various MPEG-7 Documents (notably C&O, Requirements, PPD, CfP and Evaluation Documents) were removed. New versions were issued for all these documents, based on work done in the Ad Hoc Group. Definitions and drawings were clarified in these documents.
The fact that MPEG-7 is an open, platform-independent standard was reconfirmed as a matter beyond all doubt.
The distinction between ‘Application Requirements’ and ‘MPEG-7 Requirements’ was made explicit.
First, it was clarified that the tests under discussion were for the evaluation of answers to the Call for Proposals, not for all experiments during the entire MPEG-7 development phase (i.e. for ‘Core Experiments’)
Long discussions took place on the way evaluation would be done. After the AHG meeting in Paris and discussion with usability experts, it became clear that evaluation in the context of applications would not be a good approach. It would leave too many questions about a) what is exactly evaluated? and b) how generic are the results?
Therefore, a different path was chosen, that concentrates on the basic functionality of the elements of the MPEG-7 Standard.
The result is that experts, against the requirements, will evaluate the DDL and DSs. For Descriptors (and for some Description Schemes) some kind of testing will be devised. This testing is entails making an assessment of how well material is retrieved based on similarity (‘query by example’). Proposers are further invited to explain and demonstrate their proposals.
It was decided to make not only the draft CfP and the PPD, but also the draft Evaluation Document publicly available - with the proper disclaimers.
A timeline was agreed; it can be found in the output documents. There will be a compulsory pre-registration in December.
UML was proposed as a ‘loose’ basis for the specification proposals. A one/two page guideline will be developed for proposers who don’t know UML. It was made clear that this does not constitute a choice with respect to the DDL.
An AHG on MPEG-7 Evaluation was put into place. It will have an AHG meeting on Saturday 10 and Sunday 11 October in Atlantic City.
During the meeting draft questionnaires were developed for the different types of submissions. There were some concerns on the length of the forms.
eXperimentation Model and Software Policy
Just as in MPEG-4, proposers will be required to donate source code for their descriptors upon acceptance of their proposal in the eXperimentation Model. The same policy and conditions as in MPEG-4 will apply.