v3.2.0

• Initial public release for VMAF v1. For more details please see the Netflix techblog here.

What's Changed

• libvmaf: add experimental VMAF_BATCH_THREADING and VMAF_PICTURE_POOL threading modes by @kylophone in #1482
• libvmaf/feature: add motion_v2 by @kylophone in #1483
• feature/cambi: port updates by @kylophone in #1434
• libvmaf/feature: port adm updates by @kylophone in #1485
• Port motion updates by @kylophone in #1486
• port speed updates by @kylophone in #1488
• feature/adm: temporarily disable adm_decouple_s123 x86 simd by @kylophone in #1489
• Fix null deref crash on prev_ref update in pure CUDA pipelines by @lawrencecurtis in #1487
• build(deps): bump actions/upload-artifact from 5 to 7 by @dependabot[bot] in #1466
• feature/cambi: add effective_eotf by @kylophone in #1496
• feature/float_adm: port several feature extractor options by @kylophone in #1500
• feature/motion: port several feature extractor options by @kylophone in #1497
• x86/adm: add fallback for 32-bit for 64-bit only intrinsics by @1480c1 in #1501
• feature/vif: port several feature extractor options by @kylophone in #1498
• libvmaf/feature: update alias map for cambi/speed by @kylophone in #1502
• libvmaf/feature: remove empty translation unit all.c by @kylophone in #1503
• libvmaf/test: port new adm/vif/speed tests by @kylophone in #1504
• feature/integer_vif: port vif_skip_scale0 by @kylophone in #1506
• libvmaf/predict: port chroma_from_luma correction by @kylophone in #1507
• x86/adm: fix compilation for 32-bit clang by @1480c1 in #1505
• python/test: add dataset reading tests for resampling, width/height, workfile_yuv_type by @christosbampis in #1509
• feature/cambi: remove redundant inline keyword from FORCE_INLINE by @kylophone in #1512
• python/test: add SubjectiveDatasetReader, SubjectiveDatasetTester and plot tests by @christosbampis in #1511
• port missing tests by @kylophone in #1513
• libvmaf/test: add test_motion_blend by @kylophone in #1515
• feature/vif: switch from precomputed to on-the-fly filter computation by @kylophone in #1514
• libvmaf/tools: add USE_DIRECT_READ, eliminate intermediate buffer and memcpy by @kylophone in #1516
• python: add fps filter support in Asset and enc params to call_vmafexec by @christosbampis in #1517
• python: add SpeedChroma and SpeedTemporal feature extractors and quality runners by @christosbampis in #1510
• python: refactor feature discovery to use defaultdict and support multiple nicknames by @christosbampis in #1518
• python/cambi: update to v0.8 with additional tests by @christosbampis in #1508
• python: bump VmafFeatureExtractor to v0.2.21, add aim/adm3/motion3 atom features by @christosbampis in #1519
• python: extend VmafexecQualityRunner FEATURES and add chroma_correction_parameter to TrainTestModel by @christosbampis in #1520
• python/test: port feature extractor tests for aim/adm3/motion3 by @christosbampis in #1521
• python: align whitespace and import order by @christosbampis in #1523
• python/test: adopt MyTestCase and update score values in test files by @christosbampis in #1524
• python/test: adopt MyTestCase and port new tests in asset, bootstrap, and local explainer test files by @christosbampis in #1525
• python/test: adopt MyTestCase and update score values in vmafexec tests by @christosbampis in #1526
• python/test: adopt MyTestCase, port new tests, and align fifo_mode in routine tests by @christosbampis in #1527
• python/test: adopt MyTestCase and reformat BD-rate test data by @christosbampis in #1528
• python: replace polling-based workfile synchronization with semaphores by @christosbampis in #1529
• resource/doc: add cambi_high_res_speedup parameter and update motion2 score by @christosbampis in #1530
• cambi speedup by @kylophone in #1522
• libvmaf/motion_v2: fix mirroring behavior, since a44e5e6 by @kylophone in #1531
• python/config: make download_reactively race-safe under parallel exec… by @nilfm99 in #1534
• batch_thread, pic_pool, direct_read by @kylophone in #1532
• libvmaf: replace integer_motion with pipelined v2 variant, rename by @kylophone in #1537
• libvmaf: add 2160p@1.5H CSF support (reuses 1080p@3H tables) by @christosbampis in #1539
• adm and speed simd by @kylophone in #1540
• libvmaf/motion: fix feature name collision with concurrent sfr/hfr mo… by @kylophone in #1542
• feature_collector: fix 16-byte leak with 3+ mounted models by @kylophone in #1543
• Add VMAF v1.0.16 SDR models, documentation, and tests by @kylophone in #1544

New Contributors

• @lawrencecurtis made their first contribution in #1487

Full Changelog: v3.1.0...v3.2.0