JVET-AD0332 AHG13/AHG3: software development related to FGC SEI message [P. de Lagrange, F. Urban (InterDigital)] [late]
See notes under session 4.12.
AHG4: Subjective quality testing and verification testing (6)
Contributions in this area were discussed at 1620–1710 on Saturday 22 April 2023 (chaired by JRO).
JVET-AD0332 AHG13/AHG3: software development related to FGC SEI message [P. de Lagrange, F. Urban (InterDigital)] [late]
This contribution describes the contents of the newly created JVET gitlab group called “JVET AHG Film Grain Technologies”, its connections with reference software, and the plans for future developments.
As announced on the JVET email reflector, following the mandates of AHG13 to “coordinate development of film grain technology software and configuration files” and to “coordinate with AHG3 for software support of the FGC SEI message”, a new group called “JVET AHG Film Grain Technologies” has been created on gitlab (https://vcgit.hhi.fraunhofer.de/jvet-ahg-fgt). This group is publicly accessible.
This contribution describes the current repositories within this group and the work in progress or planned.
Software developed in this group can be released independently from HM or VTM revisions, and could be merged into HM and VTM when relevant and mature enough.
Three publicly accessible repositories have been created:
- VFGS (which includes support for AR model but is not in VTM/HM)
- VTM (fork of jvet/VVCSoftware_VTM)
- HM (fork of jvet/HM)
In the contribution,
- It is recommended to continue developing FGC SEI message related software in the jvet-ahg-fgt gitlab group, merge it to the reference software when relevant and mature, and consider extra tools that enable testing and usage of film grain technologies.
- It is also recommended to study with AHG9 and AHG3 if all software related to VSEI could go into a separate repository, used as a git submodule in both HM and VTM, so that code duplication is avoided and separate versioning could be possible. This approach is currently used for SADL library in the NNVC software.
For the second recommendation above, this has nothing to do with FG synthesis, and was not discussed in this context. The first recommendation was generally agreed.
It was commented that duplication of code is undesirable. Tools that could be operated independent from HM/VTM could nevertheless become part of reference software if mature enough (which would be possible as far as they stick to the SW licensing rules). Nothing was to be decided at this meeting.