Tags

Tags give the ability to mark specific points in history as being important

  • research-alt-v1-anchor_r4.0

    0418e06c · Enable more intra mode search · 
    research anchor v1 release 4

Contains the following changes on top of v1 release 3:
- CWG-E191: Filling residual pixels outside the actual frame border
- CWG-E171: Enable more intra mode research

  • research-v9.0.0

    6fa0d27c · Fix TCQ assertion failure in windows VS · 
    Research-v9.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
CWG-E103: Move max frame width/height,color_config() forward
CWG-E156: CCSO with filter shapes
CWG-E149: MHCCP using 4 parameters
CWG-E168: Sub-pixel block vector for intraBC mode
CWG-E157: Align Deblocking Filter Boundaries With AVM Partitioning Scheme
CWG-E176: CCSO improvements
CWG-E182: Simplification of IBP DC mode
CWG-E153: Simplification for AIMC
CWG-E180: Context model optimization
CWG-E194: Secondary transforms with flexible support regions (Test 3) + 
CWG-E124: Complexity Reduction for Secondary Transforms (method 4 on + method 1 off by default)
CWG-E160: Simplification of Warp Reference List (WRL)
CWG-E140: Unify parameter computation in least square fitting
CWG-E159: DRL Improvements and Reorder Control Flag
CWG-E197: Bugfix for the signaling of transform partition types
CWG-E203: Improved Frame Context Finalization for Tile-Based Coding
CWG-E239: Adaptation rate improvement with CDF propagation
CWG-E192: Explicitly Signaled Frame-Level Non-separable Wiener Filters for Chroma
CWG-E119: Memory Reduction for RefMV Bank
CWG-E193: On Precision of Inverse DDT
CWG-E210: Context Reduction for Rectangle Partition Type
CWG-E227: Optimization of Frame Level Syntax Elements Signaling
CWG-E236: Extensive warp prediction filters
CWG-E215: Data-driven Intra Prediction
CWG-E225: Wedge Mode Simplifications
CWG-E238: Improved Partition Signaling at Frame Boundary and RDO Fixes
CWG-E235: Warp-delta mode improvement
CWG-E211: Simplification for Sub-block motion refinement
CWG-E245: BAWP Block adaptive local weighted prediction
CWG-E228: Support TIP in Low Delay
CWG-E205: A Simplification for Chroma Transform (Disabled by default)
CWG-E209 + E233: Improved signaling for compound inter prediction mode
CWG-E188: Enhanced frame context initialization
CWG-E151: IST improvement + CWG-E142: Non-normative speedup for IST
CWG-E204: Memory Bandwidth Improvement for MV Refinement Methods +
CWG-E226: 4x4 Warp Mode Constraints
CWG-E092: Trellis Coded Quantization (8-state frame-adaptive)

Encoder-only Changes
--------------------
CWG-E171: Intra mode search adjustment #
CWG-E186: Improve CDEF frame-level enabling decision
CWG-E207: Encoder and decoder optimizations
CWG-E191: Outside-frame-border block prediction and residual handling #
CWG-E221: Improvement for deblocking filter (part 1)

Code Cleanups / Improvements
----------------------------
Fix context memory analyzer
Enable 4:2:2 and 4:4:4 chroma formats in AVM Analyzer
Mismatch debug tool: increase max frame stride and height
CWG-E120: Integrate the ParaKit training software to AVM
CWG-E237: Fixes for across scale inter prediction

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/12#tab-issues

  • research-v9.0.0-rc1

    379b6174 · Increase number of shards of address to 34 · 
    Release candidate #1 for research-v9.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
CWG-E103: Move max frame width/height,color_config() forward
CWG-E156: CCSO with filter shapes
CWG-E149: MHCCP using 4 parameters
CWG-E168: Sub-pixel block vector for intraBC mode
CWG-E157: Align Deblocking Filter Boundaries With AVM Partitioning Scheme
CWG-E176: CCSO improvements
CWG-E182: Simplification of IBP DC mode
CWG-E153: Simplification for AIMC
CWG-E180: Context model optimization
CWG-E194: Secondary transforms with flexible support regions (Test 3) + 
CWG-E124: Complexity Reduction for Secondary Transforms (method 4 on + method 1 off by default)
CWG-E160: Simplification of Warp Reference List (WRL)
CWG-E140: Unify parameter computation in least square fitting
CWG-E159: DRL Improvements and Reorder Control Flag
CWG-E197: Bugfix for the signaling of transform partition types
CWG-E203: Improved Frame Context Finalization for Tile-Based Coding
CWG-E239: Adaptation rate improvement with CDF propagation
CWG-E192: Explicitly Signaled Frame-Level Non-separable Wiener Filters for Chroma
CWG-E119: Memory Reduction for RefMV Bank
CWG-E193: On Precision of Inverse DDT
CWG-E210: Context Reduction for Rectangle Partition Type
CWG-E227: Optimization of Frame Level Syntax Elements Signaling
CWG-E236: Extensive warp prediction filters
CWG-E215: Data-driven Intra Prediction
CWG-E225: Wedge Mode Simplifications
CWG-E238: Improved Partition Signaling at Frame Boundary and RDO Fixes
CWG-E235: Warp-delta mode improvement
CWG-E211: Simplification for Sub-block motion refinement
CWG-E245: BAWP Block adaptive local weighted prediction
CWG-E228: Support TIP in Low Delay
CWG-E205: A Simplification for Chroma Transform (Disabled by default)
CWG-E209 + E233: Improved signaling for compound inter prediction mode
CWG-E188: Enhanced frame context initialization
CWG-E151: IST improvement + CWG-E142: Non-normative speedup for IST
CWG-E204: Memory Bandwidth Improvement for MV Refinement Methods +
CWG-E226: 4x4 Warp Mode Constraints
CWG-E092: Trellis Coded Quantization (8-state frame-adaptive)

Encoder-only Changes
--------------------
CWG-E171: Intra mode search adjustment #
CWG-E186: Improve CDEF frame-level enabling decision
CWG-E207: Encoder and decoder optimizations
CWG-E191: Outside-frame-border block prediction and residual handling #
CWG-E221: Improvement for deblocking filter (part 1)

Code Cleanups / Improvements
----------------------------
Fix context memory analyzer
Enable 4:2:2 and 4:4:4 chroma formats in AVM Analyzer
Mismatch debug tool: increase max frame stride and height
CWG-E120: Integrate the ParaKit training software to AVM
CWG-E237: Fixes for across scale inter prediction

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/12#tab-issues

  • research-v8.1.0

    94e97f34 · Fix the nightly test failure for sequence level down-sample filter selection · 
    Minor release including following on top of v8.0.0.

Candidate Tools
---------------
- CWG-E103: Move max frame width/height,color_config() forward

Code Cleanups / Improvements
----------------------------
- Fix context memory analyzer

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/11#tab-issues

  • research-v8.0.0

    8e0b4695 · CWG-E135 TFlite local dependencies · 
    research-v8.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-E099: Further simplify and unify the DRL and the WRL
- CWG_E084: Context model optimization
- CWG-E098: Optimize the signaling of TIP direct output mode
- CWG-E107: Enlarge max inter sdp block size to 64x64
- CWG-E112: Supports compound prediction for additional blocks
- CWG-E105: Improvements/fixes in CCTX, TX partition type in WAIP, and REFINEMV
- CWG-E109: Transform type flexibility improvements
- CWG-E134: Improvements of RefMV Bank and Warp Parameter Bank
- CWG-E106: Data-Driven Transforms for Inter Coding
- CWG-E108: Fixing line-buffer and division operation issues in the CfL mode
- CWG-E133: Simplification of Skip Mode
- CWG-E122: Enabling Intra-BC for natural content
- CWG-E136: Reducing the number of partition contexts
- CWG-E132: Further Memory Reduction for Intra Bi-Prediction (overrides previous
  proposal CWG-E094)
- CWG-E139: Improved TMVP motion vector from wedge mode blocks
- CWG-E110: Pixel-Classified Explicitly Signaled Frame-Level Non-separable
  Wiener Filters
- CWG-E115: Simplified motion trajectory tracking
- CWG-E141: Unrestricted Partitioning for 1:4 Coding Blocks
- CWG-E005: Coefficient coding improvements (Part2)
- CWG-E125: Simplifications for MHCCP
- CWG-E123: Refinements for Probability Adjustment Rate Adaptation (PARA) in
  Entropy Coding

Encoder-only Changes
--------------------
- CWG-E104: Encoder mode/transform search optimizations
- CWG-E137: Encoder optimizations at speed 0 and 1
- CWG-E135: Encoder speed up using ML-based Partition Pruning, inter frames

Code Cleanups / Improvements
----------------------------
- TFLite: Bump up version to 2.17 and enable XNNPack
- CWG-E148 Research v8.0.0 anchor: Removal and Merging of Compile Time Flags

Gitlab CI Updates
-----------------
- Gitlab CI: add build check for 2 tools that are off by default.

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/10#tab-issues

  • research-v8.0.0-rc1

    0b3d929b · Enable CWG-E135, ML based partition pruning in compile time. · 
    Release candidate #1 for research-v8.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-E099: Further simplify and unify the DRL and the WRL
- CWG_E084: Context model optimization
- CWG-E098: Optimize the signaling of TIP direct output mode
- CWG-E107: Enlarge max inter sdp block size to 64x64
- CWG-E112: Supports compound prediction for additional blocks
- CWG-E105: Improvements/fixes in CCTX, TX partition type in WAIP, and REFINEMV
- CWG-E109: Transform type flexibility improvements
- CWG-E134: Improvements of RefMV Bank and Warp Parameter Bank
- CWG-E106: Data-Driven Transforms for Inter Coding
- CWG-E108: Fixing line-buffer and division operation issues in the CfL mode
- CWG-E133: Simplification of Skip Mode
- CWG-E122: Enabling Intra-BC for natural content
- CWG-E136: Reducing the number of partition contexts
- CWG-E132: Further Memory Reduction for Intra Bi-Prediction (overrides previous
  proposal CWG-E094)
- CWG-E139: Improved TMVP motion vector from wedge mode blocks
- CWG-E110: Pixel-Classified Explicitly Signaled Frame-Level Non-separable
  Wiener Filters
- CWG-E115: Simplified motion trajectory tracking
- CWG-E141: Unrestricted Partitioning for 1:4 Coding Blocks
- CWG-E005: Coefficient coding improvements (Part2)
- CWG-E125: Simplifications for MHCCP
- CWG-E123: Refinements for Probability Adjustment Rate Adaptation (PARA) in
  Entropy Coding

Encoder-only Changes
--------------------
- CWG-E104: Encoder mode/transform search optimizations
- CWG-E137: Encoder optimizations at speed 0 and 1
- CWG-E135: Encoder speed up using ML-based Partition Pruning, inter frames

Code Cleanups / Improvements
----------------------------
- TFLite: Bump up version to 2.17 and enable XNNPack

Gitlab CI Updates
-----------------
- Gitlab CI: add build check for 2 tools that are off by default.

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/10#tab-issues

  • research-v7.0.1

    6518d8e8 · Gitlab CI: add build check for 2 tools that are off by default. · 
    Minor release including following bug fixes on top of v7.0.0.

- Data race in aom_highbd_upsampled_pred_sse2:
  https://gitlab.com/AOMediaCodec/avm/-/issues/399
- Compile failure with CONFIG_ML_PART_SPLIT on:
  https://gitlab.com/AOMediaCodec/avm/-/issues/401
- Compile failure with CONFIG_BAWP_ACROSS_SCALES_FIX on:
  https://gitlab.com/AOMediaCodec/avm/-/issues/400

  • research-v7.0.0

    700b1c2b · Move allocation of pred buffer in upsampled_pref_error() · 
    research-v7.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-D181: Fix on BAWP across scales prediction
            (Disabled by default)
- CWG-D171: Low Complexity Chroma Coding
- CWG-D172: Partitioning for 1:4 Coding Blocks
- CWG-D192: Probability Model Training for Entropy Coding
- CWG-E007: merge the prediction steps for TIP reference mode
- CWG-E014: MVD coding
- CWG-E009: Quick fix of CWG-D149 and CWG-D166 integration
- CWG-D146: Quantization Matrix Simplification
- CWG-E005: Coefficient coding improvements
- CWG-E016: Unification of Cross-Plane Wiener Filters +
- CWG-E048: Optimizations to Non-Separable Wiener Filter
- CWG-E021: Simplification of Motion Vector Prediction List Construction
- CWG-E062: On block size restriction for Chroma-from-Luma prediction modes
- CWG-E049: Improvements on Identity Transform Coding
- CWG-E039: Wide Angular Intra Prediction
- CWG-E006: Linear intra prediction with matched reference templates
            (Global IBC only)
- CWG-E067: Improvement of TMVP
- CWG-E031: Bilinear interpolation for DAMR and OPFL
- CWG-E066: Addressing latency issue in SDP
- CWG-E060: On removing blocking artifacts in sub-prediction blocks
- CWG-E029: Signaling Bugfix of CCSO with Band Offset Only
- CWG-D178: Improved Lossless Coding for AVM
- CWG-E052: Complexity Reduction of Optical Flow and Affine Motion Refinement
            (aspects 2, 3 and 5)
- CWG-E055: Transform partition improvements
- CWG-E056: Enhancement of the frame output process (Overlay frames fix only)
            (overlay frames disabled by default)
- CWG-D185: Extended MV Refinement Search
- CWG-E030: Downsampling for Subblock-Based MV Refinement Search
- CWG-E065: Subblock-Level Padding for Subblock MV Refinement
- CWG-E019: Simplification of Motion Field Hole Filling
            (Aspect #1: change hole filling to 128x128 based)
- CWG-E038: On parsing dependency in intra transform signaling
- CWG-D191: Inter IST
- CWG-E023: SDP extension

Encoder-only Changes
--------------------
- CWG-E028: Context used in rate estimation fix
- CWG-E044: Disable Filter-Intra Prediction Mode
- Encoder-only Improvements for prune_ref_frames speed feature
- CWG-E070: Encoder speed up using ML-based Partition Pruning
            (Disabled by default)

Code Cleanups / Improvements
----------------------------
- CWG-E010: AVM Analyzer
- Accounting/throughput code: support for computing CDF switch counts
- Encoder Memory Reductions (Bit-exact)
- Encoder Speed Optimizations (Bit-exact)

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/9#tab-issues

Gitlab CI Updates
-----------------
- Gitlab CI: Enable assertions

  • research-v7.0.0-rc1

    985127d9 · CWG-E023 (SDP extension) · 
    Release candidate #1 for research-v7.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-D181: Fix on BAWP across scales prediction
            (Disabled by default)
- CWG-D171: Low Complexity Chroma Coding
- CWG-D172: Partitioning for 1:4 Coding Blocks
- CWG-D192: Probability Model Training for Entropy Coding
- CWG-E007: merge the prediction steps for TIP reference mode
- CWG-E014: MVD coding
- CWG-E009: Quick fix of CWG-D149 and CWG-D166 integration
- CWG-D146: Quantization Matrix Simplification
- CWG-E005: Coefficient coding improvements
- CWG-E016: Unification of Cross-Plane Wiener Filters +
- CWG-E048: Optimizations to Non-Separable Wiener Filter
- CWG-E021: Simplification of Motion Vector Prediction List Construction
- CWG-E062: On block size restriction for Chroma-from-Luma prediction modes
- CWG-E049: Improvements on Identity Transform Coding
- CWG-E039: Wide Angular Intra Prediction
- CWG-E006: Linear intra prediction with matched reference templates
            (Global IBC only)
- CWG-E067: Improvement of TMVP
- CWG-E031: Bilinear interpolation for DAMR and OPFL
- CWG-E066: Addressing latency issue in SDP
- CWG-E060: On removing blocking artifacts in sub-prediction blocks
- CWG-E029: Signaling Bugfix of CCSO with Band Offset Only
- CWG-D178: Improved Lossless Coding for AVM
- CWG-E052: Complexity Reduction of Optical Flow and Affine Motion Refinement
            (aspects 2, 3 and 5)
- CWG-E055: Transform partition improvements
- CWG-E056: Enhancement of the frame output process (Overlay frames fix only)
            (overlay frames disabled by default)
- CWG-D185: Extended MV Refinement Search
- CWG-E030: Downsampling for Subblock-Based MV Refinement Search
- CWG-E065: Subblock-Level Padding for Subblock MV Refinement
- CWG-E019: Simplification of Motion Field Hole Filling
            (Aspect #1: change hole filling to 128x128 based)
- CWG-E038: On parsing dependency in intra transform signaling
- CWG-D191: Inter IST
- CWG-E023: SDP extension

Encoder-only Changes
--------------------
- CWG-E028: Context used in rate estimation fix
- CWG-E044: Disable Filter-Intra Prediction Mode
- Encoder-only Improvements for prune_ref_frames speed feature
- CWG-E070: Encoder speed up using ML-based Partition Pruning
            (Disabled by default)

Code Cleanups / Improvements
----------------------------
- CWG-E010: AVM Analyzer
- Accounting/throughput code: support for computing CDF switch counts
- Encoder Memory Reductions (Bit-exact)
- Encoder Speed Optimizations (Bit-exact)

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/9#tab-issues

  • research-v6.1.0

    de986bff · av1_determine_sc_tools_with_encoding: reorder initialization · 
    research-v6.1.0 minor release.

Includes the following changes since research-v6.0.0 anchor:

Candidate Tools
---------------
- CWG-D181: Fix on BAWP across scales prediction (off by default)
- CWG-D171: Low Complexity Chroma Coding
- CWG-D172: Partitioning for 1:4 Coding Blocks
- CWG-D192: Probability Model Training for Entropy Coding

Code Cleanups / Improvements
----------------------------
- CWG-E010: AVM Analyzer
- Misc memory reductions

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/8#tab-issues

  • research-alt-v1-anchor_r3.0

    9ee9eb7f · Add support for PSNR calculation based on stream bit-depth · 
    research anchor v1 release 3

Contains the following changes on top of v1 release 2:
- CWG-D098: Temporal filtering improvement
- CWG-D082: Screen content options determination improvement
- CWG-D088: Coding 8-bit Video using 10-bit
- Bugfixes to support CTCv6 and other fixes.

  • research-v6.0.0

    b22a8f6c · Add support for PSNR calculation based on stream bit-depth · 
    research-v6.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-D014: Frame output process based on order hint information (Enable by
            default)
- CWG-D096: Explicit signaling of a secondary transform set
- CWG-D117: Improvement of transform partition context modeling
- CWG-D009: ADST types replacement
- CWG-D048: Explicit BAWP (Disabled by default)
- CWG-D078: BLOCK_256 Partitions
- CWG-D068: Rectangular 1:8 and 1:16 blocks
- CWG-D089: Improved Chroma Transform Coefficient Coding
- CWG-D049: Joint Chroma-luma optimization for BAWP
- CWG-D094: Adaptive Edge Classifier for CCSO
- CWG-D070: On Block Vector Predictor List Size of IntraBC
- CWG-D115: PARA: Probability Adaptation Rate Adjustment for Entropy Coding
- CWG-D138: Retrained IST Kernels
- CWG-D132: Improvement of Inter-intra Mode of AVM
- CWG-D145: Selecting and signaling of the primary reference frame
- CWG-D067: Improved warp motion of AVM - Aspect 3
- CWG-D044: Refresh frame flags signaling (Disabled by default)
- CWG-D149: Improvement in Entropy Context Model
- CWG-D159: R-D Optimized IST Set Selection
- CWG-D042: Chroma Prediction Mode Signaling
- CWG-D072: Improvements of Skip Mode
- CWG-D153: Improvements of Compound Prediction with Same Reference Frames
            (aspects 1,2,3 and 6)
- CWG-D063: Frame Level Motion Vector for TIP Direct Mode
- CWG-D143: CCSO with Joint Enabling Flag
- CWG-D165: Disable OBMC
- CWG-D118: Multi-hypothesis Cross Component Intra Prediction
- CWG-D161: EOB Position Coding for Luma
- CWG-D144: CCSO with Band Offset Only Option
- CWG-D097: Improvement of cross-component Wiener filter for chroma components
- CWG-D086: Context modeling for transform block zero flag signaling
- CWG-D155: TIP Mode with Implicit Quantization Indices
- CWG-D166: Compound warp prediction for AVM (Extend WARP_CAUSAL to NEW_NEWMV)
- CWG-D114: Palette Line Copy
- CWG-D084: Decoder Side Affine Motion Refinement (including baseline change)

Encoder-only Changes
--------------------
- CWG-D090: Mean-removed SSE (Disabled by default)
- CWG-D098: Temporal filtering improvement
- CWG-D082: Screen content options determination improvement
- CWG-D088: Coding 8-bit Video using 10-bit
- CWG-D093: Enable transform partition search for inter blocks of size 16x16 and
            smaller
- CWG-D136: Optical flow based MV Search

Code Cleanups / Improvements
----------------------------
- CWG-D123: AVM Inspection and Analysis (core tooling)
- CWG-D160: Removal and Merging of Compile Time Flags
- Heap Memory for encoder and decoder
- CWG-D182: An Entropy Coding Memory Analyzer for AVM

Gitlab CI Updates
-----------------
- Upgrade LLVM Version to 15
- Upgrade GCC Version to 12

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/7#tab-issues

  • research-v6.0.0-rc1

    e9e5e10d · Fix multiple issues in D118 (MHCCP) · 
    Release candidate #1 for research-v6.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-D014: Frame output process based on order hint information (Enable by
            default)
- CWG-D096: Explicit signaling of a secondary transform set
- CWG-D117: Improvement of transform partition context modeling
- CWG-D009: ADST types replacement
- CWG-D048: Explicit BAWP (Disabled by default)
- CWG-D078: BLOCK_256 Partitions
- CWG-D068: Rectangular 1:8 and 1:16 blocks
- CWG-D089: Improved Chroma Transform Coefficient Coding
- CWG-D049: Joint Chroma-luma optimization for BAWP
- CWG-D094: Adaptive Edge Classifier for CCSO
- CWG-D070: On Block Vector Predictor List Size of IntraBC
- CWG-D115: PARA: Probability Adaptation Rate Adjustment for Entropy Coding
- CWG-D138: Retrained IST Kernels
- CWG-D132: Improvement of Inter-intra Mode of AVM
- CWG-D145: Selecting and signaling of the primary reference frame
- CWG-D067: Improved warp motion of AVM - Aspect 3
- CWG-D044: Refresh frame flags signaling (Disabled by default)
- CWG-D149: Improvement in Entropy Context Model
- CWG-D159: R-D Optimized IST Set Selection
- CWG-D042: Chroma Prediction Mode Signaling
- CWG-D072: Improvements of Skip Mode
- CWG-D153: Improvements of Compound Prediction with Same Reference Frames
            (aspects 1,2,3 and 6)
- CWG-D063: Frame Level Motion Vector for TIP Direct Mode
- CWG-D143: CCSO with Joint Enabling Flag
- CWG-D165: Disable OBMC
- CWG-D118: Multi-hypothesis Cross Component Intra Prediction
- CWG-D161: EOB Position Coding for Luma
- CWG-D144: CCSO with Band Offset Only Option
- CWG-D097: Improvement of cross-component Wiener filter for chroma components
- CWG-D086: Context modeling for transform block zero flag signaling
- CWG-D155: TIP Mode with Implicit Quantization Indices
- CWG-D166: Compound warp prediction for AVM (Extend WARP_CAUSAL to NEW_NEWMV)
- CWG-D114: Palette Line Copy
- CWG-D084: Decoder Side Affine Motion Refinement (including baseline change)

Encoder-only Changes
--------------------
- CWG-D090: Mean-removed SSE (Disabled by default)
- CWG-D098: Temporal filtering improvement
- CWG-D082: Screen content options determination improvement
- CWG-D088: Coding 8-bit Video using 10-bit
- CWG-D093: Enable transform partition search for inter blocks of size 16x16 and
            smaller
- CWG-D136: Optical flow based MV Search

Code Cleanups / Improvements
----------------------------
- CWG-D123: AVM Inspection and Analysis (core tooling)

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/7#tab-issues

  • research-v5.0.0

    ca0e5fd3 · Fix compile flag names in CONFIG_ENTROPY_STATS · 
    research-v5.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-D026: Temporal motion projection improvements
- CWG-D016: Improved angular intra prediction modes
- CWG-D020: Improved Signaling Scheme for ERP
- CWG-D045: Prediction Mode Dependent Skip Transform
- CWG-D023: DC-based TX Restriction and EOB Alignment
- CWG-D047: Add by-pass check in deblocking strength decision
- CWG-D027: compound weighted prediction
- CWG-D043: Bug Fix for H-Partitions
- CWG-D010 + CWG-D052: Non-separable wiener filter improvements
- CWG_D041: Extended Intra Prediction Angles
- CWG-D035: Uneven 4-way partitions
- CWG-C034: Intra Directional Interpolation Filter
- CWG-D059: Explicit Frame Distance Calculation
- CWG-D034: Improved Global Motion
- Disable global motion by default (related to CWG-D034)
- Signal CFL Downsampling Filter in Sequence Header (Decision from CWG
  Call on 2023/06/16)
- CWG-D015: Wedge Blending Improvement
- CWG-D074: SMVP Fix
- CWG-D029: On downsampling filter in CfL mode
- CWG-D067: Improved warp motion of AVM (Aspects 1 and 2)
- CWG-D031: Sub-block based MV refinement
- CWG-D028: On downsampling filter in the CfL mode for the 4:2:2 chroma
  format
- CWG-D060: Separate DRL for Compound Prediction
- CWG-D071: Implicit Masked Blending Mode / Boundary-aware Compound
  Prediction
- CWG-D046: Blending Modes For Intra Prediction
- CWG-D061: Improve Encoding Speed of ERP
- CWG-D014: (Disabled by default) Frame Output Process Based On Order
  Hint

Code Cleanups
-------------
- Inspection / Accounting improvements
- CWG-D037, #3: Combine multiple compile-time flags into one

Gitlab CI Updates
-----------------
- CWG-D037, #6: More helpful error messages for failed CI jobs, which
  point to relevant wiki.
- Speed-up Per-MR CI
- Speed-up Nightly CI
- Add Decode Checks in CI

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/5#tab-issues

  • research-v5.0.0-rc1

    66bf5d0d · Disable CONFIG_OUTPUT_FRAME_BASED_ON_ORDER_HINT by default. · 
    Release candidate #1 for research-v5.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-D026: Temporal motion projection improvements
- CWG-D016: Improved angular intra prediction modes
- CWG-D020: Improved Signaling Scheme for ERP
- CWG-D045: Prediction Mode Dependent Skip Transform
- CWG-D023: DC-based TX Restriction and EOB Alignment
- CWG-D047: Add by-pass check in deblocking strength decision
- CWG-D027: compound weighted prediction
- CWG-D043: Bug Fix for H-Partitions
- CWG-D010 + CWG-D052: Non-separable wiener filter improvements
- CWG_D041: Extended Intra Prediction Angles
- CWG-D035: Uneven 4-way partitions
- CWG-C034: Intra Directional Interpolation Filter
- CWG-D059: Explicit Frame Distance Calculation
- CWG-D034: Improved Global Motion
- Disable global motion by default (related to CWG-D034)
- Signal CFL Downsampling Filter in Sequence Header (Decision from CWG
  Call on 2023/06/16)
- CWG-D015: Wedge Blending Improvement
- CWG-D074: SMVP Fix
- CWG-D029: On downsampling filter in CfL mode
- CWG-D067: Improved warp motion of AVM (Aspects 1 and 2)
- CWG-D031: Sub-block based MV refinement
- CWG-D028: On downsampling filter in the CfL mode for the 4:2:2 chroma
  format
- CWG-D060: Separate DRL for Compound Prediction
- CWG-D071: Implicit Masked Blending Mode / Boundary-aware Compound
  Prediction
- CWG-D046: Blending Modes For Intra Prediction
- CWG-D061: Improve Encoding Speed of ERP
- CWG-D014: (Disabled by default) Frame Output Process Based On Order
  Hint

Code Cleanups
-------------
- Inspection / Accounting improvements

Gitlab CI Updates
-----------------
- CWG-D037, #6: More helpful error messages for failed CI jobs, which
  point to relevant wiki.
- Speed-up Per-MR CI
- Speed-up Nightly CI
- Add Decode Checks in CI

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/5#tab-issues

  • research-v4.0.0

    ae397e62 · Fix size of warp ref list · 
    research-v4.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-C020: Flex MV precisions
- CWG-C030: Frame-level on/off flag for CDEF
- CWG-C044: Improved CfL prediction mode
- CWG-C043: MVP list improvements
- CWG-C035: Optical flow refinement for TIP frames
- CWG-C042: Allow Same-Ref Compound
- CWG-C050: Extended Warp Prediction tools
- CWG-C033: Skip mode DRL list with reference index
- CWG-C053: Improved JMVD coding
- CWG-C059 / CWG-C060: Coefficient Coding and New TX Sets
- CWG-C063: Improved TMVP candidates
- CWG-C061: WRL for warp-delta mode
- CWG-C064: A SIMD Implementation for CCSO Filter
- CWG-C054: On downsampling filter in CfL mode
- CWG-C071: Subblock MV from Local Warp
- CWG-C039: Parity hiding for transform block
- CWG-C075: Entropy coding context modeling improvement
- CWG-C076: Inter mode contexts
- CWG-C069: Block adaptive local weighted prediction
- CWG-B010: New transform partitions (enable by default)
- CWG-C068: Cross-chroma Component Transforms
- CWG-C080: Improved Warp
- CWG-C073: A Simplified Transform Set and Configurable Transform Partitioning
- CWG-C087: On Local Warp Parsing Dependency
- CWG-C089: On warp motion of AVM
- CWG-C086: Prediction Enhancement Filter
- CWG-B084: Extended Recursive Partitions
- CWG-C079: Entropy Bypass Improvement
- CWG-C092: Wedge Mode Extensions
- CWG-C016: Extensions to Loop-Restoration
- CWG-C080: ERP Amendment: H-Partitions

Code Cleanups
-------------
- Remove all `CONVERT_TO_BYTEPTR` and `CONVERT_TO_SHORTPTR` from codebase
- CWG-C091: Removal of Compile Time Flags

Gitlab CI Updates
-----------------
- Nightly test to ensure parallel / serial GOP encodes match

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/4#tab-issues

  • research-v4.0.0-rc1

    40b7086d · Compiler warning fixes related to NEW_TX_PARTITION · 
    Release candidate #1 for research-v4.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-C020: Flex MV precisions
- CWG-C030: Frame-level on/off flag for CDEF
- CWG-C044: Improved CfL prediction mode
- CWG-C043: MVP list improvements
- CWG-C035: Optical flow refinement for TIP frames
- CWG-C042: Allow Same-Ref Compound
- CWG-C050: Extended Warp Prediction tools
- CWG-C033: Skip mode DRL list with reference index
- CWG-C053: Improved JMVD coding
- CWG-C059 / CWG-C060: Coefficient Coding and New TX Sets
- CWG-C063: Improved TMVP candidates
- CWG-C061: WRL for warp-delta mode
- CWG-C064: A SIMD Implementation for CCSO Filter
- CWG-C054: On downsampling filter in CfL mode
- CWG-C071: Subblock MV from Local Warp
- CWG-C039: Parity hiding for transform block
- CWG-C075: Entropy coding context modeling improvement
- CWG-C076: Inter mode contexts
- CWG-C069: Block adaptive local weighted prediction
- CWG-B010: New transform partitions (enable by default)
- CWG-C068: Cross-chroma Component Transforms
- CWG-C080: Improved Warp
- CWG-C073: A Simplified Transform Set and Configurable Transform Partitioning
- CWG-C087: On Local Warp Parsing Dependency
- CWG-C089: On warp motion of AVM
- CWG-C086: Prediction Enhancement Filter
- CWG-B084: Extended Recursive Partitions
- CWG-C079: Entropy Bypass Improvement
- CWG-C092: Wedge Mode Extensions
- CWG-C016: Extensions to Loop-Restoration
- CWG-C080: ERP Amendment: H-Partitions

Code Cleanups
-------------
- Remove all `CONVERT_TO_BYTEPTR` and `CONVERT_TO_SHORTPTR` from codebase
- CWG-C091: Removal of Compile Time Flags

Gitlab CI Updates
-----------------
- Nightly test to ensure parallel / serial GOP encodes match

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/4#tab-issues

  • research-v3.1.0

    2fc12fd8 · Change variable name for MVD class offset · 
    Includes the following changes since research-v3.0.0 anchor:

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/milestones/3#tab-issues

Added feature
--------
-  Frame hash metadata is added for checking the enc/dec mismatch.

  • research-v3.0.0

    945b8a28 · Disable CONFIG_MV_SEARCH_RANGE · 
    research-v3.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-B023 REF_MV_BANK
- CWG-B041 Optical Flow MV Refinement
- CWG-B079 Interpolation Filter Consistency Fix in Optical Flow MV Refinement
- CWG-B056 Intra Bi-Prediction
- CWG-B068 Adaptive Intra Mode Coding (AIMC)
- CWG-B053 Complexity-scalable MVP
- CWG-B076 Scan order alignment and bugfix for IST
- CWG-B065 Modified context derivation
- CWG-B098 Improvements to IST
- CWG-B081 Search range extension for IntraBC
- CWG-B099 Improved CCSO with luma extension and band feature
- CWG-B092 Advanced MVD coding
- CWG-C002 Forward Skip Coding for Prediction Residuals (FSC)
- CWG-C011 Improved AMVD
- CWG-C005 Improved Palette Coding
- CWG-B091 New Reference Signaling
- CWG-C019 SKIP_MODE_ENHANCEMENT
- CWG-C014 New Deblocking Filter
- CWG-C025 BVP_IMPROVEMENT
- CWG-C007 Temporal Interpolated Prediction Mode
- CWG-C031: Rate cost update for block vector

Code Cleanups
-------------
- CWG-B086 Removal of DIST_WTD_COMP code
- CWG-B086 Removal of DUAL_FILTER code
- CWG-B086 Removal of SVC Features
- CWG-B086 Removal of some compile-time flags
- CWG-B086 Removal of 8-bit pipeline from encoder and decoder

Gitlab CI Updates
-----------------
- Upgraded GCC to version 11
- Upgraded LLVM to version 12
- Added developer wiki for reproducing CI failures

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/issues/?sort=created_date&state=closed&milestone_title=research-v3.0.0%20anchor%20release

  • research-v3.0.0-rc1

    91e91dc4 · update Tensorflow to 2.6.0 · 
    Release candidate #1 for research-v3.0.0 anchor.

Includes the following since the previous anchor:

Candidate Tools
---------------
- CWG-B023 REF_MV_BANK
- CWG-B041 Optical Flow MV Refinement
- CWG-B079 Interpolation Filter Consistency Fix in Optical Flow MV Refinement
- CWG-B056 Intra Bi-Prediction
- CWG-B068 Adaptive Intra Mode Coding (AIMC)
- CWG-B053 Complexity-scalable MVP
- CWG-B076 Scan order alignment and bugfix for IST
- CWG-B065 Modified context derivation
- CWG-B098 Improvements to IST
- CWG-B081 Search range extension for IntraBC
- CWG-B099 Improved CCSO with luma extension and band feature
- CWG-B092 Advanced MVD coding
- CWG-C002 Forward Skip Coding for Prediction Residuals (FSC)
- CWG-C011 Improved AMVD
- CWG-C005 Improved Palette Coding
- CWG-B091 New Reference Signaling
- CWG-C019 SKIP_MODE_ENHANCEMENT
- CWG-C014 New Deblocking Filter
- CWG-C025 BVP_IMPROVEMENT
- CWG-C007 Temporal Interpolated Prediction Mode
- CWG-C031: Rate cost update for block vector

Code Cleanups
-------------
- CWG-B086 Removal of DIST_WTD_COMP code
- CWG-B086 Removal of DUAL_FILTER code
- CWG-B086 Removal of SVC Features
- CWG-B086 Removal of some compile-time flags

Gitlab CI Updates
-----------------
- Upgraded GCC to version 11
- Upgraded LLVM to version 12
- Developer Wiki for Reproducing CI Failures

Bugfixes
--------
https://gitlab.com/AOMediaCodec/avm/-/issues/?sort=created_date&state=closed&milestone_title=research-v3.0.0%20anchor%20release

Open issues
-----------
https://gitlab.com/AOMediaCodec/avm/-/issues/?sort=created_date&state=opened&milestone_title=research-v3.0.0%20anchor%20release