- 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