Closing plenary topics
A closing plenary of the Video subgroup was held Thursday 16:00–18:00, and continued Friday 8:00-9:00. Discussions are recorded in this section. It was confirmed that some breakout activities such as finalization in editing of documents and reporting for document approval in other subgroups could continue after the closing Friday morning, and no objections were raised on this.
Detailed reports were given by the BoGs.
VCC:
-
Presentation of evaluation by M Wien
-
Confirms initial assessment made Thu morning
-
A report (non public) will be issued with 4 weeks editing period
CDVS:
-
CE2: Agreement was reached that the scope is restricted to improvements based on the SFV aggregation component
-
It was confirmed that the data used for training of the global descriptor will be delivered by Jan 24, otherwise the new TM9 and the Study of CD will not include the modification of the global descriptor from m32261.
-
Some other changes of TM are non-normative
-
Other changes in the Study text should be of purely editorial nature, if e.g. the editors come to the conclusion that technical change such as re-organization of the bitstream would be beneficial, this should be proposed as input to the next meeting.
-
Close CE3. The deviation between different detectors is too large, and it is impractical to describe the necessary commonality of the extraction process in a unique way without explicitly describing one specific process. Nevertheless, more consideration is necessary which variations of the normative extraction process could still be called “CDVS compliant”, and if it is at all possible to specify conformance. This will be investigated in the AHG.
-
In CE2, request results at two different points of false positive rate (1% and 0.1%).
IVC:
-
Both new elements (res dep interpolation filter and prediction block size dependent ABT) are confirmed to be based on expired patents. New versions of m32173 and m32174 were uploaded which are documenting this. These will only be included in ITM8, not in the WD.
WVC:
-
It is suggested to issue the FDIS and DoC by the next meeting.
VCB:
-
Further review of the draft was performed in a breakout following the video plenary (chaired by Qunshan Gu)
The following items were raised during the plenary:
-
Experts from HanYang University commented that the document m32443, comments 24, 25 and 26 have not been considered in the latest version of the draft but need to be carefully considered. It is the opinion of the contributors of m32443 and the editors that this can be incorporated in the editing period.
-
It is necessary at various places to remove the implementation-related aspects in the spec and replace with more neutral decoding process descriptions.
-
It was asked whether there is need to define profiles and levels? It is agreed to postpone this to a later stage, as it did not seem necessary at the WD stage.
-
How about conformance? This will be done after profiles and levels are defined.
-
There is excessive use of pseudo code in the document rather than textual description of decoder operation, eg. in subclause 4.3.2.2.
-
There is no overall bitsteam structure specified in the document The editor thinks it is adequate, however it will be tried to improve.
-
The syntax is imprecise and incomplete. Eg. Table 2.4 10-4 on p. 28.
-
Table numbers need fixing.
-
Various places have formuation of writing in the first person, probably some remainder from software code comments.
The following reasons were raised against promoting to CD
-
lack of editorial quality
-
incomplete specification, syntax and semantics not complete
-
imprecise description of decoding process
-
Too few experts have collaborated so far in improving it
On the other hand, in this specific case there is no doubt that the software of the VCB codec is mature, interoperability of decoders on different platforms is proven, such that it should give a clear reference about what should be in the text description. Promoting it to CD would increase the attention and enforce NBs to consider it more seriously than if it stays as WD. However, regarding the final timeline, there would be no difference whether the CD is issued now or by the next meeting. With the current status, an editing period would be needed, such that the ballot could anyway not close before the next meeting.
It was therefore agreed to issue a Request for subdivision (14496-31) with the following timeline: CD 14/04, DIS 14/07, FDIS 15/02.
The document with test conditions for possible VCB extensions was also reviewed, the sequence Party Scene was not in the initial draft but shall be included.
Issuing a WD3 was approved, and an editing period of 2 weeks was granted for improvements.
Dostları ilə paylaş: |