Refactoring:
- Lift check for exec existence during program load.
- Refactor psnr, ssim, ms_ssim and vmaf_feature to call ExternalProgramCaller.
- Refactor feature/Makefile to make executables depend on libvmaf.a.
- Refactor wrapper/Makefile to include additional objs in libvmaf.a but exclude main.o.
- Remove ar -d command after removing main.o from libvmaf.a.
New features:
- Generalize read_dataset.
- Update default Asset resampling method to bicubic (#116).
- Extend ffmpeg2vmaf script to allow ref/dis input to be YUV (#118).
- Improve README.md (#121).
Fixed bugs:
- Temporary fix Visual Studio builds (#112).
- Avoid unnecessary dependency on matplotlib in run_vmaf (#114).
- Remove unneeded dependencies in Dockerfile, fixes #115 (#117).
- MinGW support (#123).
- Change compute_vmaf() interface to return an error code instead of throw an error #124 (#126).
Refactoring:
- Refactor NorefExecutorMixin to eliminate repeated codes.
- Refactor C code: get rid of unused double functions; uniformly use read_frame callback function to void repeated code;
- Add strip option to Makefile.
New features:
- Update Asset class: add copy functions to Asset; add ref/dis_yuv_type; deprecate yuv_type; add ref/dis_start_sec;
- Update subjective models: add confidence interval to subjective model parameters; refactor MLE model and make subclasses; add run_subj command line.
- Recommend pip, add ffmpeg2vmaf info and reorganize prerequisite installation (#88).
- Reduce sleep time in parallel_map.
- Add library interface for VMAF (#90).
- Add VisualStudio2015 support (#92).
- Add example of image dataset notyuv.
- Add pkgconfig file and changed Makefile.
- Add VmafPhoneQualityRunner class.
- Add DMOS_MLE_CO subjective model.
Fixed bugs:
- Update RegressionMixin to handle KFLK exception for dicitonary-style dataset.
- Fix Makefile fedora libptools issue. (#98)
Refactoring:
- Deprecate run_executors_in_parallel.
- Refactor NorefFeatureExtractor into NorefExecutorMixin so that it can be used for all executors.
- Add abstract methods to some base classes.
New features:
- Add ST-RRED runner (StrredQualityRunner), based on "Video Quality Assessment by Reduced Reference Spatio-Temporal Entropic Differencing", by R. Soundararaajan, A. Bovik.
- Add start/end frame support for Executor.
New features:
- Refactor to replace config.ROOT with config.VmafConfig.
New features:
- Generalize Result and FileSystemResultStore to allow None values.
Tasks:
- Refactor to prepare for pypi packaging.
New features:
- Updated VMAF model to version v0.6.1. Changes include: 1) added a custom model for cellular phone screen viewing; 2) trained using new dataset, covering more difficult content; 3) elementary metric fixes: ADM behavior at near-black frames, motion behavior at scene boundaries; 4) compressed quality score range by 20% to accommodate higher dynamic range; 5) Use MLE instead of DMOS as subjective model.
Fixed bugs:
- Replace subprocess.call with run_process (checking return value).
New features:
- Add command line ffmpeg2vmaf, which takes encoded videos as input.
New features:
- Allow processing non-YUV input videos.
New features:
- Add STRRED runner.
New features:
- Allow specifying crop and pad parameter in dataset files.
Fixed bugs:
- Replace pathos with custom function for parallel executor running.
Fixed bugs:
- Fix command line run_testing issue. Add command line test cases.
New features:
- Speed up VMAF convolution operation by AVX.
Fixed bugs:
- Fix vmafossexec memory leakage.
New features:
- Add enable_transform_score option to VmafQualityRunner, VmafossExecQualityRunner.
Fixed bugs:
- Fix a bug in DatasetReader.to_aggregated_dataset_file.
New features:
- Add Travis continuous integration.
New features:
- Add implementation of KFLK - quality metric evaluation method based on AUC. Refer to: L. Krasula, K. Fliegel, P. Le Callet, M.Klima, "On the accuracy of objective image and video quality models: New methodology for performance evaluation", QoMEX 2016.
New features:
- Add options to use custom subjective models in run_vmaf_training and run_testing commands.
New features:
- Add DatasetReader and subclasses; add SubjectiveModel and subclasses.
New features:
- Add quality runners for each individual VMAF elementary metrics.
Fixed bugs:
- Issue #36: SSIM and MS-SSIM sometimes get negative values.
New features:
- Add Xcode project support.
- Add more pooling options (median, percx) to CLIs.
New features:
- Add support for docker usage (#30).
Fixed bugs:
- Issue #29: Make ptools build under Fedora.
New features:
- Generalize dataset format to allow per-content YUV format.
Fixed bugs:
- Make ptools work under Mac OS.
- Update SklearnRandomForestTrainTestModel test with sklearn 0.18.
New features:
- Update command lines run_vmaf, run_psnr, run_vmaf_in_batch, run_cleaning_cache, run_vmaf_training and run_testing.
New features:
- Update wrapper/vmafossexec: 1) it now takes pkl model file as input, so that slopes/intercepts are no longer hard-coded; 2) it now takes multiple YUV input formats; 3) add flag to enable/disable VMAF score clipping at 0/100; 4) allow customly running PSNR/SSIM/MS-SSIM; 5) allow customly outputing XML/JSON
- Add SSIM/MS-SSIM option in run_testing.
Fixed bugs:
- Move VmafQualityRunnerWithLocalExplainer to quality_runner_adhoc to resolve multiple instances of VMAF found when calling QualityRunner.find_subclass.
New features:
- Add custom_clip_0to1 to TrainTestModel.
New features:
- Generalize read_dataset to allow specifying width, height and resampling method on which to calculate quality.
- Add bicubic to SUPPORTED_RESAMPLING_TYPES for Asset.
- Update Asset rule with resampling_type in str to avoid duplicates in data store.
New features:
- Update VmafFeatureExtractor to 0.2.2b with scaled ADM features exposed (adm_scale0-3).
New features:
- Add DisYUVRawVideoExtractor and related classes.
- Add NeuralNetworkTrainTestModel base class that integrates TensorFlow.
- Add example class ToddNoiseClassifierTrainTestModel.
New features:
- Add LocalExplainer class.
- Add show_local_explanation option to run_vmaf script.
Fixed bugs:
- Fix a series of numerical issues in VMAF features, increment VmafFeatureExtractor version number.
- Retrain VmafQualityRunner after feature update, increment version number.
New features:
- Add base class NorefFeatureExtractor for any feature extractor that do not use a reference video.
- Add MomentNorefFeatureExtractor subclassing NorefFeatureExtractor as an example implementation.
New features:
- Refactor feature code to expose ssim/ms-ssim, speed up ssim/ms-ssim.
Fixed bugs:
- Fix feature while looping by moving feof to after read_image.
- Fix issue #2 use hashed string for log filename and result filename to avoid file names getting too long.
New features:
- Add SsimFeatureExtractor and MsSsimFeatureExtractor with intermediate features (luminence, contrast, structure).