Back to Search Document details
13th Meeting: Marrakech, January 2019 2019-01-16 16:41
BoG report on neural networks for video coding
Abstract
This contribution provides the report of the BoG on neural networks (NN) for video coding, especially for neural network based loop filter. An information report (JVET-M0691) is discussed in this BoG, and then the CE plan.
JVET-M0904 BoG report on neural networks for video coding [Y. Li, S. Liu]

This BoG report was presented in Track B on Thursday 17 January 1500-1815 (chaired by JRO).

This contribution provides the report of the BoG on neural networks (NN) for video coding, especially for neural network-based loop filtering. An information report (JVET-M0691) was discussed in this BoG, and then a CE plan.

The BoG recommended the following plan of a CE:

  • Divide the 6 NN based loop filter methods into two categories and build two sub CE tests.

SubCE1 is the test for sequence-adaptive (two-pass) method, SubCE2 is the test for sequence-independent (one-pass) method.

  • Investigate the impact of NN filter position in the filter chain.

Due to there being many test cases, only choose some cases to test, as defined below.

    • For SubCE1:

anchor: DBF + SAO + ALF

case 1a: DBF + SAO + ALF + NN filter, which is the case currently subCE1 contributions used.

case 2a: NN filter. (optional test)

    • For SubCE2:

anchor: DBF + SAO + ALF

case 1a: DBF + NN filter + SAO + ALF, which is the case that most of the contributions used.

case 2a: NN filter. (optional test)

case 3a: NN filter + ALF. (optional test)

  • Investigate the benefit of the CTU/block level NN filter adaptive on/off.

Define some cases to test. To test this part, other parts should be fixed, e.g., using case 1a as the position set and case 1c as the QP set.

case 1b: make the NN filter adaptive on/off at CTU level.

case 2b: always use NN filter in one slice. (optional test)

  • Investigate the performance of the NN filter when the test QP is not the same as the training QP.

Define some cases to test. To test this part, one should fix using case 1a and case 1b.

Using CTC QP (22, 27, 32, 37) for training and get the NN parameter.

case 1c: based on the NN parameter, use CTC QP for testing

case 2c: based on the NN parameter, use CTC QP + 2 for testing (24, 29, 34, 39) (optional test)

case 3c: based on the NN parameter, use CTC QP - 2 for testing (20, 27, 30, 35) (optional test)

  • More detailed assignment of CE can be seen in tables provided in the BoG report.

The BoG recommended the following to be discussed.

  • For subCE1, due to the online training, the NN parameter is not fixed. Thus, only providing the code in the inference stage is insufficient for others to use. (Although the parameter training/generation process is encoder only)

The proponent of Intel is willing to provide the training method, but not before the next meeting. The use of Tensorflow or other third-party framework in the common test conditions should be discuss in the track. Answer from the track: It may be useful if it is precisely described by which parameters a package like tensorflow is operated. An external optimization framework does not necessarily need to be part of our software package at the current stage of investigation.

  • For subCE1, training process need GPU (using CPU is 10x slower), but the inference stage/second pass only use CPU. The training process is separated from the second pass. Currently, the encoding time listed in the table only includes the coding time for the second pass, and the training time is reported separately. Does it need to add the training time and second coding time for reporting? Or whether there is another better solution? Answer from the track: Should be reported, but not necessarily included in encoding time, as this would mix inhomogeneous computing platforms, and conclusions about the possibility of doing this in a meaningful amount of time can be drawn from separate reporting

An initial draft of a CE description was also included in the BoG report.

It was planned to run some of the tests only for short parts (e.g. one Intra refresh period). It was questioned whether this would give sufficient information; it was, however, planned to proceed like this for initial tests to identify configurations which will then be finally tested with entire sequences.

For the resulting CE plan, see JVET-M1033.

JVET-M0904 BoG report on neural networks for video coding [Y. Li, S. Liu]

See section 6.17.

List of actions taken affecting Draft 3 of VVC, VTM 3, and 360Lib

The following is a summary, in the form of a brief list, of the actions taken at the meeting that affect the text of the VVC draft text, VTM or 360Lib description. Both technical and editorial issues are included. This list is provided only as a summary – details of specific actions are noted elsewhere in this report and the list provided here may not be complete and correct. The listing of a document number only indicates that the document is related, not that it was adopted in whole or in part.

Category

Motivation

Modification

AI
BD-R Y

RA
BD-R Y

Document

Decision

In-loop filters

ALF

Fix

pcm_loop_filter_disabled_flag for ALF

0.0%

0.0%

JVET-M0277

Decision (BF/text): Include disabling ALF as a third loop filter when the pcm_loop_filter_disabled_flag is set, as suggested in JVET-M0277

Deblocking

Subjective quality

Long deblocking

0.1%

0.0%

JVET-M0471

Decision: Adopt JVET-M0471, version 11.1.8 (specification text available in v2 upload, but needs another small modification for restriction of line buffer, was shortly reviewed in track B Thursday 17 January 1330), pending on confirmation from the viewing, and the more detailed report on complexity impact

Deblocking

Subjective quality

Deblocking of CIIP boundaries

JVET-M0908

combination of JVET-M0103 and JVET-M0294

Reconstruction

Coding efficiency

Picture reconstructon with mapping

-0.9%

-1.3%

JVET-M0427

Decision: Adopt (modified as noted).

Intra

Prediction mode

Coding efficiency

Intra subpartitions

-0.6%

-0.3%

JVET-M0102

Decision: Adopt 1.1.1 proposal, pending the provision of text and its review.

CCLM

Coding efficiency (HDR)

Modified CCLM downsampling filter

0.0%

0.0%

JVET-M0142

Decision: Adopt 2.4.c with a high-level flag to switch between two chroma format type optimizations (pending test results for applying the type 2 scheme to type 0 content).

CCLM

Simplification

Table reduction in CCLM modelling

0.0%

0.0%

JVET-M0064

Decision: Adopted

Prediction

Cleanup

Harmonize the ref sample filtering

JVET-M0095

Editorial action item: Agreed

PDPC

Simplification

Simplified linear interpolation

0.0%

0.0%

JVET-M0238

Decision: Adopted

CPR

Coding efficiency (SCC)

Reference sample memory reuse

JVET-M0407

Decision: Adopt JVET-M0407 (variant a)

CPR

Cleanup

CPR signalling - interaction with inter tools

JVET-M0483

Decision: Adopt JVET-M0483 (text in zip v4, “r1”, probably needs some more update along the lines above)

Transform

MTS

Complexity reduction

Fast DST-7/DCT-8

0.0%

0.0%

JVET-M0497

Decision (complexity reduction): Adopt CE6-2.3a.

MTS

Complexity reduction

32-length DST-7/DCT-8 using zero-out

0.1%

0.0%

JVET-M0297

Decision: Adopt JVET-M0297 Test 2.

MTS/TS

Simplification + coding efficiency (SCC)

Unifed MTS/TS syntax + TS up to 32x32

0.0%

0.0%

JVET-M0464

Decision: Adopt transform skip up to 32x32, with associated syntax approach in JVET-M0464 using tu_mts_idx. And enable TS in CTC for all test sequences.

TU partitioning

Coding efficiency

Sub-block Transform (SBT) for inter blocks

-0.5%

JVET-M0140

Decision (coding efficiency): Adopt CE6-4.1a (pending the test results of avoiding 32-point DST). The same high-level flag as used for CE6-3.1b is to be used to determine whether DCT2 is used always or not and also applies to CE3-1.1.1.

DCT2

Fix (editorial)

Ticket #135: Downsampling of the DCT2 transform matrix should be separate horizontally and vertically (not assumed square)

JVET-M0002

Decision (BF editorial): The downsampling of the DCT2 transform matrix should be separate horizontally and vertically (not assumed square) for #135.

Quantization

QP

Fix

Bug fix for quantization group QP signalling

JVET-M0113
JVET-M0188

Decision: Adopt (text in JVET-M0113).

QP

Fix

QP Prediction fix for parallel encoding

JVET-M0685

Decision: Initialize QP from the bottom left CU of the above CTU row when decoding the first CU of a CTU on the left edge of a tile group

Scaling

Fix

Modified dequantization scaling for TS

JVET-M0119

Decision (bug fix): Adopt.

CABAC

Contexts

Complexity reduction

Reduce merge idx ctx coded bins

0.0%

JVET-M0381

Decision: Adopt JVET-M0381 Test 2.2.2a (reducing number of context coded bins in affine merge). Text is available with the contribution.

Contexts

Coding efficiency

1 additional context for pred_mode_flag

-0.1%

JVET-M0502

Decision: all the suggested adoptions are confirmed by trackB. (Method 2)

Engine

Coding efficiency

Probability estimation

-1.0%

-1.0%

JVET-M0453

Decision: adopt “5.1.13* +new init from 5.1.2”

Residual Coding

Complexity reduction

rem_abs_gt3_flag in first coding pass

-0.1%

-0.1%

JVET-M0173

Decision (complexity reduction): Adopt CE7.4.

Residual Coding

Fix

Limited EGk for abs_rem/ dec_abs_level

0.0%

0.0%

JVET-M0470

Decision (BF): Adopt JVET-M0470.

Residual Coding

Fix

Last position coding for large block-size transforms

JVET-M0251
JVET-M0257

Decision (BF): Adopt JVET-M0251/JVET-M0257 (software from JVET-M0257).

Inter

SBTMVP

Complexity reduction

Only using left neighbour for SbTMVP

0.0%

JVET-M0273

Decision: all the suggested adoptions are confirmed by trackB.

Affine

Coding efficiency

AMVR for affine

-0.2%

JVET-M0246

Decision: Adopt JVET-M0246 (Test 2.1.2), extending AMVR to affine. Use the AMVR high level flag for disabling both “normal” AMVR and “affine” AMVR.

Affine merge

Cleanup

Affine sub-block MV clipping

0.0%

JVET-M0145

Decision: all the suggested adoptions are confirmed by trackB.

Affine merge

Complexity reduction

Remove MV comparison

0.0%

JVET-M0166

Decision: all the suggested adoptions are confirmed by trackB. JVET-M0166 (change 2)/ JVET-M0228 (modification 1)/ JVET-M0477(change 2)

Merge

Fix

Fix for cu_cbf when merge

JVET-M0361

Decision: Agreed in Track A review – this was just an error in drafting the text.

Merge

Complexity reduction

Parallel processing for merge mode

0.0%

JVET-M0170

Decision: Adopt JVET-M0170 (type 2, draft text “…type2sharing” of Jan. 12)

Merge

Complexity reduction

Pairwise Average Candidate Reduction

0.0%

JVET-M0193

Decision: all the suggested adoptions are confirmed by trackB.

Merge/AMVP

Complexity reduction

Rounding before any MV pruning

0.0%

JVET-M0281

Decision: Adopt JVET-M0281 (subtest 4.1.5a)

AMVP

Complexity reduction

MVP candidate list generation for AMVP

0.0%

JVET-M0117

Decision: all the suggested adoptions are confirmed by trackB.

HMVP

Cleanup

Reduce HMVP number from 6 to 5

0.0%

JVET-M0436

Decision: all the suggested adoptions are confirmed by trackB.

HMVP

Complexity reduction

HMVP and parallel processing with tiles

0.0%

JVET-M0300

Decision: all the suggested adoptions are confirmed by trackB.

HMVP

Cleanup

GBi weight is also stored in HMVP

0.0%

JVET-M0264

Decision: all the suggested adoptions are confirmed by trackB.

HMVP

Simplification

HMVP candidate pruning

0.0%

JVET-M0126

Decision: Adopt JVET-M0126 version 4.1.2.4 (text is available, but needs to be reduced reflect that only this aspect is changed.

BDOF

Complexity reduction

Integer positions in extended region

0.1%

JVET-M0487

Decision: Adopt JVET-M0487 (solution 9.1.1.b)

BDOF

Fix

Generalization of BDOF bit-depth

0.0%

JVET-M0063

Decision (BF): Adopt JVET-M0063.

MMVD

Coding efficiency (SCC)

MMVD w/o Fractional Distances for SCC

-0.1%

JVET-M0255

Adopt the approach of not signalling the triangular prediction mode flag in cases where the combination is not allowed (MMVD, CIIP) – various contributions on that, gives 0.07%.

MMVD

Cleanup

Harmonize MV scaling

0.0%

JVET-M0068

Decision: all the suggested adoptions are confirmed by trackB.

MMVD

Fixes/cleanup

M0068+Forbid 4*4 bi + align SW with WD

0.0%

JVET-M0171

Decision: all the suggested adoptions are confirmed by trackB.

MVD

Coding efficiency

Symmetrical MVD coding for L0 to L1

-0.3%

JVET-M0444

Decision: all the suggested adoptions are confirmed by trackB.

WP

fixes/cleanup

Disable GBI signalling when WP is enabled

JVET-M0111

Decision: all the suggested adoptions are confirmed by trackB.

MV

cleanup

Clip MVs to 18 bits

0.0%

JVET-M0479

Decision: all the suggested adoptions are confirmed by trackB.

MV

Complexity reduction

TMVP Storage Reduction

0.0%

JVET-M0512

Decision: Adopt JVET-M0512 second aspect as described in notes

MV

Complexity reduction

Sub-block MV derivation for chroma

0.0%

JVET-M0192

Decision: all the suggested adoptions are confirmed by trackB.

DMVR

Coding efficiency

DMVR

-1.1%

JVET-M0147

Decision: Adopt JVET-M0147 with SAD cost function, and without the MVD based early termination check.

Triangular

Simplification

-0.1%

JVET-M0118

1. Adopt syntax change of JVET-M0118: Same as in JVET-M0185, JVET-M0190, JVET-M207 (test 1), JVET-M0216 (the first aspect), JVET-M0234 (change corresponding to the result table 7 and 8), JVET-M0317 (section 2.2), JVET-M0328 (test E). This does not signal the triangular prediction mode flag in cases where the combination is not allowed (MMVD, CIIP). Approx. 0.07% rate reduction

Triangular

Simplification

0.0%

JVET-M0328

2. Adopt using only weight group (second weight as in JVET-M0328). The weighting is currently dependent on comparing MVs of the two partitions, loss approx. 0.01% for RA, 0.03% for LB, also simplifies the spec.

Triangular

Simplification

0.0%

JVET-M0883

4. Adopt signalling change of triangular merging candidate which does not need LUT (JVET-M0883). This is a combination from various proposals. Does not change merge list construction, simplifies the specification, and gives tiny gain (0.01% both for RA and LB)

Partitioning

Cleanup

inferred QT split to avoid 32x128/128x32 partitions at picture boundaries

0.1%

JVET-M0905
JVET-M0888
JVET-M0446

Decision (cleanup/consistency): Adopt inferred QT split to avoid 32x128/128x32 partitions at picture boundaries (0.06% penalty in RA configuration).

Cleanup

Split-first signalling for partitioning

JVET-M0421

Decision: adopt

High-level syntax

Picture referencing

Simplification

RPL-based reference picture management

JVET-M0128

Decision: Adopt with modifications as described in notes

JVET-M0101

Replace IRAP_NUT with 3 new NAL unit types: IDR_W_RADL, IDR_N_LP, CRA_NUT

JVET-M0101

Add external means flag HandleCraAsCvsStartFlag

JVET-M0101

Add a NUT value for STSA and AUD

JVET-M0101

Add sps_max_sub_layers_minus1 syntax element to SPS, and decoding process in 8.1.1, 8.1.2 and 8.1.3 of JVET-M0101

JVET-M0101

Add profile_tier_level( ) syntax structure which includes sub layer level idc

JVET-M0101

Add general_non_packed_constraint_flag with semantics as in JVET-M0101

JVET-M0101

Add the temporal scalability sub-bitstream extraction process in JVET-M0101

JVET-M0415

Change the sps_ref_wraparound_offset to sps_ref_wraparound_offset_minus1 and changing the units to be MinCbSizeY, subject to review by the 360° BoG

JVET-M0451

Add 7 new constraint flags corresponding to VVC WD 3 tools as described in JVET-M0451

JVET-M0128

Add reference picture signalling from JVET-M0128 (basic text version) – modified after Track A discussion as described in the notes for that document

Decision: Add RASL and RADL NUTs

Tiles and WPP

JVET-M0853

Decision: Adopted, with constraints as in notes

JVET-M0132

Decision: Adopt an adaptation parameter set (APS) to carry ALF parameters. The tile group header contains an aps_id which is conditionally present when ALF is enabled. The APS contains an aps_id and the ALF parameters. A new NUT value is assigned for APS (from JVET-M0132). For the CTC, we will just use aps_id = 0 and send the APS with each picture. For now, the range of APS ID values will be 0..31 and APSs can be shared across pictures (and can be different in different tile groups within a picture). The ID value should be fixed-length coded when present. ID values cannot be re-used with different content within the same picture.

Add loop_filter_across_tile_group_enabled_flag to the PPS

JVET-M0160

Decision: Adopted (confirmed Thursday morning plenary).

Software & CTC

Intra prediction

Fix

Ticket #132: Mismatch between spec and software in the order of syntax elements

JVET-M0002

Decision (SW): A software fix was needed for #132.

Motion search

Coding efficiency (SCC)

Hash-based Motion Search for SCC

0.0%

0.0%

JVET-M0253

Decision (SW): Adopt JVET-M0253

Subblock merge

Fix

Mismatch between text specification and reference software on ATMVP candidate derivation when CPR is enabled

0.0%

0.0%

JVET-M0409

Decision (SW/BF): Adopt JVET-M0409 (align software with text)

MV

Fix

Clean-up on MV Rounding

0.0%

0.0%

JVET-M0265

Decision (BF/consistency): Adopt rounding away from zero for MV averages.

MMVD

Coding efficiency

MMVD improvement

-0.1%

JVET-M0823

Decision (SW): Adopt JVET-M0823, also CTC

PCM

Coding efficiency/fixes

SAO/ALF

0.0%

0.0%

JVET-M0277

Decision (BF/SW): Disable SAO and ALF in chroma part when dual tree is used and the flag is set, and disable ALF in luma part when dual tree is used and the flag is set, as suggested in JVET-M0277

Partitioning

Coding efficientcy

Encoder optimization for RDO

0.0%

0.0%

JVET-M0428

Decision (SW): Adopt JVET-M0428, not for CTC (-0.6% -0.7%)

Tiles

Encoder motion constraints for MCTSs

JVET-M0445

Decision (SW): Adopted

AMVR

Affine motion estimation for affine AMVR+C94

-0.1%

JVET-M0247

Decision: all the suggested adoptions are confirmed by trackB.

Affine merge

Increasing the amount of RD checking for affine merge mode coding

-0.1%

JVET-M0839

Decision: all the suggested adoptions are confirmed by trackB.

Analysis

Functionality

Enhancement of the tool to measure the memory bandwidth in VTM

JVET-M0864

Decision (SW): Adopt JVET-M0864 memory bandwidth analysis method.

Functionality

VTM transcoding capabilities

JVET-M0055

Decision (SW): adopt, usage should be documented in VTM SW package

Quantization

Functionality

Clean-up of perceptually optimized QP adaptation

JVET-M0091

Decision (SW): Adopt

Quantization

Fix (CTC)

Rebalance luma-vs.-chroma QP setting

-1.0%

JVET-M0090

Decision (CTC): Adopt

Rate control

Functionality

Bug fix for rate control

JVET-M0511

Decision (SW): Adopt

Functionality

Quality dependency factor based rate control

JVET-M0600

Decision (SW): Adopt

JVET-M0452

Adopt Hemisphere CMP and Hemisphere EAC projection formats (from JVET-M0452)

JVET-M0368

Modify chroma sample location in blending process for PHEC (from JVET-M0368) – basically a bug fix

Overall (CTC) ex Class F & SCC

-2.4%

-6.2%

Project planning

Core experiment planning

Core experiment plans were established, and initial drafts of the CE plans were reviewed in the closing sessions.

Drafting of specification text, encoder algorithm descriptions, and software

The following agreement has been established: the editorial team has the discretion to not integrate recorded adoptions for which the available text is grossly inadequate (and cannot be fixed with a reasonable degree of effort), if such a situation hypothetically arises. In such an event, the text would record the intent expressed by the committee without including a full integration of the available inadequate text.

Plans for improved efficiency and contribution consideration

The group considered it important to have the full design of proposals documented to enable proper study.

Adoptions need to be based on properly drafted working draft text (on normative elements) and HM encoder algorithm descriptions – relative to the existing drafts. Proposal contributions should also provide a software implementation (or at least such software should be made available for study and testing by other participants at the meeting, and software must be made available to cross-checkers in EEs).

Suggestions for future meetings included the following generally-supported principles:

  • No review of normative contributions without draft specification text
  • VTM algorithm description text is strongly encouraged for non-normative contributions
  • Early upload deadline to enable substantial study prior to the meeting
  • Using a clock timer to ensure efficient proposal presentations (5 min) and discussions

The document upload deadline for the next meeting was planned to be Tuesday 12 March 2019.

As general guidance, it was suggested to avoid usage of company names in document titles, software modules etc., and not to describe a technology by using a company name.

General issues for experiments

It was emphasized during the opening plenary on January 9 that those rules which had been set up or refined during the 12th meeting should be observed. In particular, for some CEs, results had been available late, and some changes in the experimental setup (particularly in CE4) had not been discussed on the JVET reflector.

Group coordinated experiments have been planned as follows:

  • “Core experiments” (CEs) are the coordinated experiments on coding tools which are deemed to be interesting but require more investigation and could potentially become part of the draft standard by the next meeting.
  • A CE is a test of a specific fully described technology in a specific agreed way. It is not a forum for thinking of new ideas (like an AHG). The CE coordinators are responsible for making sure that the CE description is complete and correct and has adequate detail. Reflector discussions about CE description clarity and other aspects of CE plans are encouraged.
  • A description of each experiment is to be approved at the meeting at which the experiment plan is established. This should include the issues that were raised by other experts when the tool was presented, e.g., interference with other tools, contribution of different elements that are part of a package, etc. The experiment description document should provide the names of individual people, not just company names.
  • Software for tools investigated in a CE will be provided in one or more separate branches of the software repository. Each CE will have a “fork” of the software, and within the CE there may be multiple branches established by the CE coordinator. The software coordinator will help coordinate the creation of these forks and branches and their naming. All JVET members have read access to the CE software branches (using shared read-only credentials; one account has been set up to use the usual shared WG 11 credentials and another is set up for SG 16 members with credentials shared in the TIES system at the location https://www.itu.int/ifa/t/2017/sg16/exchange/wp3/q06/vceg_account.txt).
  • During the experiment, revisions of the experiment plans can be made, but not substantial changes to the proposed technology.
  • The CE description must match the CE testing that is done. The CE description needs to be revised if there has been some change of plans.
  • The CE summary report must describe any changes that were made in the process of finalizing the CE plan.
  • By the next meeting it is expected that at least one independent cross-checker will report a detailed analysis of each proposed feature that has been tested and confirm that the implementation is correct. Commentary on the potential benefits and disadvantages of the proposed technology in cross-checking reports is highly encouraged. Having multiple cross-checking reports is also highly encouraged (especially if the cross-checking involves more than confirmation of correct test results). The reports of cross-checking activities may (and generally should) be integrated into the CE report rather than submitted as separate documents.

It is possible to define sub-experiments within particular CEs, for example designated as CE9-2.3a, CE9-2.3b, etc., for subtests 2.3a and 2.3b in CE9.

As a general rule, it was agreed that each CE should be run under the same testing conditions using one software codebase, which should be based on the group test model software codebase. An experiment is not to be established as a CE unless there is access given to the participants in (any part of) the CE to the software used to perform the experiments.

The general agreed common conditions for single-layer coding efficiency experiments are described in the output document JVET-M1010.

Experiment descriptions should be written in a way such that it is understood as a JVET output document (written from an objective “third party perspective”, not a proponent perspective – e.g. not referring to methods as “improved”, “optimized”, etc.). The experiment descriptions should generally not express opinions or suggest conclusions – rather, they should just describe what technology will be tested, how it will be tested, who will participate, etc. Responsibilities for contributions to CE work should identify individuals in addition to company names.

CE descriptions contain a basic description of the technology under test, but should not contain excessively verbose descriptions of a technology (at least not unless the technology is not adequately documented elsewhere). Instead, the CE descriptions should refer to the relevant proposal contributions for any necessary further detail. However, the complete detail of what technology will be tested must be available – either in the CE description itself or in documents that are referenced in the CE description that are also available in the JVET document archive.

Any technology must have at least one cross-check partner to establish a CE – a single proponent is not enough. It is highly desirable have more than just one proponent and one cross-checker.

Some agreements relating to CE activities were established as follows:

  • Only qualified JVET members can be identified participants in a CE.
  • Participation in a CE is possible without a commitment of submitting an input document to the next meeting. Participation is requested by contacting the CE coordinator.
  • All software, results, and documents produced in the CE should be announced and made available to JVET in a timely manner.
  • All substantial communications about a CE, other than logistics arrangements, exchange of data, minor refinement of the test plans, and preparation of documents shall be conducted on the main JVET reflector. In the case that large amounts of data are to be distributed is recommended to send an announcement to the JVET reflector without attaching the materials, and send the materials to those who have requested it directly, or provide a link to it, or upload the data as an input contribution to the next meeting.

General timeline for CEs

T1= 3 weeks after the JVET meeting: To revise the CE description and refine questions to be answered. Questions should be discussed and agreed on JVET reflector. Any changes of planned tests after this time need to be announced and discussed on the JVET reflector.

T2 = Test model software release + 2 weeks or 4 March, whichever is earlier: Integration of all tools into a separate CE branch of the VTM is completed and announced to JVET reflector.

  • Initial study by cross-checkers can begin.
  • Proponents may continue to modify the software in this branch until T3
  • 3rd parties are encouraged to study and make contributions to the next meeting with proposed changes

T3: 3 weeks before the next JVET meeting or T2 + 1 week, whichever is later: Any changes to the CE test branches of the software must be frozen, so the cross-checkers can know exactly what they are cross-checking. A software version tag should be created at this time and announced on the JVET reflector. The name of the cross-checkers and list of specific tests for each tool under study in the CE plan description by this time. Full test results must be provided at this time (at least for proposals targeting to be promoted to the draft standard at the next meeting).

CE reports may contain additional information about tests of straightforwared combinations of the identified technologies. Such supplemental testing needs to be clearly identified in the report if it was not part of the CE plan.

New branches may be created which combine two or more tools included in the CE document or the VTM (as applicable).

It is not necessary to formally name cross-checkers in the initial version of the CE description document. To adopt a proposed feature at the next meeting, we would like see comprehensive cross-checking done, with analysis that the description matches the software, and recommendation of value of the tool given tradeoffs.

The establishment of a CE does not indicate that a proposed technology is mature for adoption or that the testing conducted in the CE is fully adequate for assessing the merits of the technology, and a favourable outcome of CE does not indicate a need for adoption of the technology.

Draft specification text shall be provided with CE input documents. Availability of spec text is important to have a detailed understanding of the technology and also to judge what its impact on the complexity of the spec will be. There must also be sufficient time to study it in detail. CE contributions without sufficiently mature draft spec text in the CE input document (available by the time of the document deadline) should not be considered for adoption. Furthermore, CE contribution documents should be complete and not make it necessary to open old documents to understand the technology.

Plans for the CEs to be conducted were established Thursday 18 January (chaired by GJS); CE plan documents were reviewed Friday 19 January (GJS & JRO).

Lists of participants in CE documents should be pruned to include only the active participants. Read access to software is available to all members.

Software development and anchor generation

The planned timeline for software releases was established as follows:

  • VTM4.0 will be released by 2019-02-11. VTM4.1 with non-CTC adoptions will be released later. (If necessary, VTM4.0 may not include final tuning of context initialization values.)
  • Further versions of VTM may be released for additional bug fixing, as appropriate.
  • Preparation of the VTM software will include immediate removal of macros that were added in the previous meeting cycle. The software coordinator has the discretion to retain some such macros.
  • Timeline of 360lib9.0: 1 week after the release of VTM4.0 (2019-02-18). Further versions may be released as appropriate for bug fixing.

Establishment of ad hoc groups

The ad hoc groups established to progress work on particular subject areas until the next meeting are described in the table below. The discussion list for all of these ad hoc groups was agreed to be the main JVET reflector (jvet@lists.rwth-aachen.de).

Title and Email Reflector

Chairs

Mtg

Project Management (AHG1)

(jvet@lists.rwth-aachen.de)

  • Coordinate overall JVET interim efforts.
  • Supervise CE and AHG studies.
  • Report on project status to JVET reflector.
  • Provide a report to next meeting on project coordination status.

J.-R. Ohm, G. J. Sullivan (co-chairs)

N

Draft text and test model algorithm description editing (AHG2)

(jvet@lists.rwth-aachen.de)

  • Produce and finalize JVET-M1001 VVC text specification draft 4.
  • Produce and finalize JVET-M1002 VVC Test Model 4 (VTM 4) Algorithm and Encoder Description.
  • Gather and address comments for refinement of these documents.
  • Coordinate with test model software development AhG to address issues relating to mismatches between software and text.

B. Bross, J. Chen (co-chairs), J. Boyce, S. Kim, S. Liu, Y. Ye (vice-chairs)

N

Test model software development (AHG3)

(jvet@lists.rwth-aachen.de)

  • Coordinate development of test model (VTM) software and associated configuration files.
  • Produce documentation of software usage for distribution with the software.
  • Discuss and make recommendations on the software development process.
  • Propose improvements to the guideline document for developments of the test model software.
  • Perform tests of VTM 4 behaviour relative to HEVC and VTM 3 using the VTM common test conditions and the multi-resolution streaming test conditions described in JVET-M0466.
  • Coordinate with AHG on Draft text and test model algorithm description editing (AHG2) to identify any mismatches between software and text, and make further updates and cleanups to the software as appropriate.
  • Coordinate with AHG6 for integration with 360lib software.

F. Bossen, X. Li, A. Norkin, K. Sühring (co-chairs)

N

Test material and visual assessment (AHG4)

(jvet@lists.rwth-aachen.de)

  • Maintain the video sequence test material database for development of the VVC standard.
  • Identify and recommend appropriate test materials for use in the development of the VVC standard.
  • Identify missing types of video material, solicit contributions, collect, and make available a variety of video sequence test material.
  • Evaluate new test sequences, particularly including the material recently submitted by the Blender Foundation / Blender Animation Studio and Twitch.
  • Propose a new structure for the test sequence repository.
  • Facilitate availability of viewing equipment and facilities arrangements for the next meeting and pre-meeting testing as feasible.

T. Suzuki (chair), V. Baroncini, R. Chernyak, P. Hanhart, A. Norkin, J. Ye (vice-chairs)

N

Memory bandwidth consumption of coding tools (AHG5)

(jvet@lists.rwth-aachen.de)

  • Develop improved software tools for measuring both average and worst case of memory bandwidth, and provide information for usage of these tools.
  • Study cache configurations for measuring decoder memory bandwidth consumption.
  • Identify coding tools in CEs and VTM with significant memory bandwidth impact.
  • Study the impact of memory bandwidth on specific application cases.

R. Hashimoto (chair), T. Ikai, X. Li, D. Luo, H. Yang, M. Zhou (vice-chairs)

N

360° video conversion software development (AHG6)

(jvet@lists.rwth-aachen.de)

  • Prepare and deliver the 360Lib-9.0 software version and common test condition configuration files according to JVET-M1012.
  • Generate CTC (PHEC) anchors and PERP results for VTM according to JVET-M1012, and finalize the reporting template for the common test conditions.
  • Produce documentation of software usage for distribution with the software.

Y. He, K. Choi (co-chairs)

N

Coding of HDR/WCG material (AHG7)

(jvet@lists.rwth-aachen.de)

  • Study and evaluate available HDR/WCG test content.
  • Study objective metrics for quality assessment of HDR/WCG material, including investigation of the correlation between subjective and objective results of the CfP responses.
  • Compare the performance of the VTM and HM for HDR/WCG content.
  • Prepare for expert viewing of HDR content at the 14th JVET meeting if feasible.
  • Coordinate implementation of HDR anchor aspects in the test model software with AHG3.
  • Study additional aspects of coding HDR/WCG content.

A. Segall (chair), E. François, W. Husak, D. Rusanovskyy (vice-chairs)

N

360° video coding tools and test conditions (AHG8)

(jvet@lists.rwth-aachen.de)

  • Study the effect on compression and subjective quality of different projections formats, resolutions, and packing layouts.
  • Discuss refinements of common test conditions, test sequences, and evaluation criteria.
  • Solicit additional test sequences, and evaluate suitability of test sequences on head-mounted displays and normal 2D displays.
  • Study coding tools dedicated to 360° video, their impact on compression, and implications to the core codec design.
  • Study the effect of viewport resolution, field of view, and viewport speed/direction on visual comfort.
  • Study complexity of GPU rendering of projection formats
  • Study syntax for signalling of projection formats, cubeface layouts, spherical rotations

J. Boyce (chair), K. Choi, P. Hanhart, J.-L. Lin (vice-chairs)

N

Neural networks in video coding (AHG9)

(jvet@lists.rwth-aachen.de)

  • Investigate the benefit of using neural networks in video compression such as CNN loop filter, intra prediction, re-sampling in adaptive resolution coding, and encoder side partition mode decisions.
  • Investigate the complexity impact of using neural networks in video compression.
  • Investigate the complexity measurement of neural network coding tools.
  • Investigate the impact of training materials on the performance of neural network coding tools.
  • Investigate the impact of the training process on performance and complexity.

S. Liu (chair), B. Choi, K. Kawamura, Y. Li, L. Wang, P. Wu, H. Yang (vice-chairs)

N

Encoding algorithm optimization (AHG10)

(jvet@lists.rwth-aachen.de)

  • Study the impact of using techniques such as GOP structures and perceptually optimized adaptive quantization for encoder optimization.
  • Study the impact of adaptive quantization on individual tools in the test model.
  • Study the quantization adaptation tool in the test model.
  • Investigate the feasibility of adding a CTC test category in which adaptive quantization is turned on.
  • Study quality metrics for measuring subjective quality using e.g. the CfP response MOS scores.
  • Investigate other methods of improving objective and/or subjective quality, including adaptive coding structures, adaptive quantization without signalling, and multi-pass encoding.
  • Study methods of rate control and their impact on performance, subjective and objective quality.

A. Duenas, A. Tourapis (co-chairs), C. Helmrich, S. Ikonin, A. Norkin, R. Sjöberg, T. Toma (vice-chairs)

N

Screen content coding (AHG11)

(jvet@lists.rwth-aachen.de)

  • Investigate coding tools targeted at screen content in terms of compression benefit and implementation complexity.
  • Identify test materials, discuss testing conditions for screen content coding, and propose associated updated common test conditions.
  • Study the impact of loop filters on screen content coding

S. Liu (chair), J. Boyce, A. Filippov, Y.-C. Sun, J. Xu, M. Zhou (vice-chairs)

N

High-level parallelism and coded picture regions (AHG12)

(jvet@lists.rwth-aachen.de)

  • Study wavefront processing including the relationship with tiles and low delay characteristics.
  • Study flexible loop filter control and tile size restriction, including identifying implications on coding tools and implementation.
  • Study flexible tile partitioning (e.g. more flexible than HEVC and tile boundaries not spanning a full picture).
  • Study support of independently coded picture regions, including easy rewriting of such regions into a conforming sub-bitstream.
  • Prepare software and configurations for the test model to facilitate parallel processing tests.
  • Study the coding efficiency impact of parallel processing and coded picture regions.

S. Deshpande (chair), M. M. Hannuksela, R. Sjöberg, R. Skupin, W. Wan, Y.-K. Wang S. Wenger (vice-chairs)

N

Tool reporting procedure (AHG13)

(jvet@lists.rwth-aachen.de)

  • Prepare output document JVET-M1005, which describes the methodology of tool-off testing and a list of tools to be tested by identified testers.
  • Provide configurations files, bitstreams, and results of tool-on/tool-off testing.
  • Use the tool usage counts and memory bandwidth usage to study the decoder complexity of features in on/off testing.
  • Prepare a report with results of the tests.

W.-J. Chien, J. Boyce (co-chairs), Y.-W. Chen, R. Chernyak, K. Choi, R. Hashimoto, Y.-W. Huang, H. Jang, S. Liu, D. Luo (vice-chairs)

N

Progressive intra refresh (AHG14)

(jvet@lists.rwth-aachen.de)

  • Define relevant test conditions for the study of progressive intra refresh for random access without intra frames
  • Update the implementation of encoder-only intra refresh in the VTM model in the AHG14 fork of the software repository.
  • Evaluate different ways to produce intra refresh within VVC and characterize their coding efficiency impact, subjective quality, and delay characteristics, including encoder-only approaches and normative approaches
  • Consider the use of constrained intra prediction and tile-based approaches
  • Study recovery point handling, including practical implementation issues and perfect-versus-approximate decoded picture recovery.
  • Consider the potential need for starting a coded video sequence without an intra picture.

J.-M. Thiesse (chair), A. Duenas, K. Kazui, R. Sjöberg, A. Tourapis (vice-chairs)

N

Bitstream decoding properties signalling (AHG15)

(jvet@lists.rwth-aachen.de)

  • Study syntax alternatives for interoperability point signalling
  • Study selection of constraint flags to be included in the VTM and their impact on syntax, semantics, and decoding process

J. Boyce (chair), J. Chen, S. Deshpande, M. Karczewicz, A. Tourapis, Y.-K. Wang, S. Wenger (vice-chairs)

N

Implementation studies (AHG16)

(jvet@lists.rwth-aachen.de)

  • Study draft and proposed coding tools to identify implementation issues relating to decoder pipelines, decoder throughput, and other aspects of implementation difficulty.
  • Solicit hardware analysis of complex tools.
  • Particularly consider intra reconstruction throughput for small blocks.
  • Provide feedback on potential solutions to address identified issues.

M. Zhou (chair), J. An, E. Chai, K. Choi, S. Ethuraman, T. Hsieh, X. Xiu (vice-chairs)

N

High-level syntax (AHG17)

(jvet@lists.rwth-aachen.de)

  • Study NAL unit header, sequence parameter set, picture parameter set, adaptation parameter set, and tile group header syntax designs
  • Study the proposed picture header designs and alternatives
  • Study reference picture buffering and list construction
  • Study random access signalling and random access approaches, including approaches with reference pictures provided by external means
  • Assist in software development and text drafting for the high-level syntax in the VVC design.

R. Sjöberg (chair), S. Deshpande, M. M. Hannuksela, R. Skupin, Y.-K. Wang, S. Wenger, H. Yu (vice-chairs)

N

Quantization control (AHG18)

(jvet@lists.rwth-aachen.de)

  • Identify methods for quantization step size control for luma and chroma, including spatially and frequency-adaptive approaches
  • Develop methods for evaluating quantization step size control operation
  • Study the impact of MTS transforms on quantization matrices and the need for default matrices
  • Study the interaction between in-loop “reshaping” and quantization step size control
  • Develop testing conditions for evaluating QP signalling improvements including rate control and perceptual optimization strategies as appropriate
  • Evaluate the performance of the current VVC QP design using the two adaptive quantization control techniques currently available in the VTM

R. Chernyak (chair), E. François, C. Helmrich, A. Segall (vice-chairs)

N

Layered coding and resolution adaptivity (AHG19)

(jvet@lists.rwth-aachen.de)

  • Study adaptive-resolution coding approaches for real-time communication, adaptive streaming, and 360-degree viewport-dependent streaming, including filters for resampling, reference picture management, and related scope and signalling
  • Study approaches for temporal scalability to avoid temporal judder when temporal scalability sub-bitstream extraction is used for achieving lower frame rate, and consider whether this should have a normative impact.
  • Identify related test conditions, test sequences, and evaluation techniques (including subjective assessment techniques)
  • Study potential approaches for support of layered coding scalability including spatial, temporal, quality, and view scalability

S. Wenger and A. Segall (co-chairs), M. M. Hannuksela, Hendry, S. McCarthy, Y.-C. Sun (vice-chairs)

N

Output documents

The following documents were agreed to be produced or endorsed as outputs of the meeting. Names recorded below indicate the editors responsible for the document production. Where applicable, dates of planned finalization and corresponding parent-body document numbers are also noted.

It was reminded that in cases where the JVET document is also made available as MPEG output document, a separate version under the MPEG document header should be generated. This version should be sent to GJS and JRO for upload.

Decisions
For the resulting CE plan, see JVET-M1033.
adopted
all the suggested adoptions are confirmed by trackB
Citation