-
Notifications
You must be signed in to change notification settings - Fork 535
/
Changelog
executable file
·128 lines (95 loc) · 5.96 KB
/
Changelog
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
Changelog
---------
Essentia 1.0:
- first public release
Essentia 1.0.1:
- fixed a lot of memory leaks
- fixed a crash in the bpm histogram descriptor
Essentia 1.0.2:
- fixed all the remaining memory leaks. Essentia is now 100% leak-free!
Essentia 1.0.3:
- extractor in windows now prints the message of the exception if any was thrown
(just crashed silently before)
- added RealAccumulator algorithm
- ValueFile can now output string and vector<string>
- cleaned examples and made sure they all work
Essentia 1.0.4:
- fixed serious bug in the extractor, ReplayGain value was always 0
- fixed crash when using deleteNetwork on some AlgorithmComposites
- added range description for parameters, also does automatic range checking when
configuring an algorithm
- added new python unit tests and streaming C++ unit tests
- did a white-box review of Essentia's most used algorithms, they now:
- have input/output/parameter names which conform better to the guidelines
- do input validation
- some descriptors output by the streaming_extractor have been fixed and now have
different values. Here is the list of changes:
- MFCC are slightly different (negligible), because the bands limits were not
correctly taken before (off by 1 error)
- Barkbands were overlapping before, not anymore so some bands are the same,
some others have smaller energy values.
differences are significative, and are more significative in lower bands
(because we have lower resolution in low frequencies)
Descriptors dependent on barkbands:
- barkbands_kurtosis -> changes a lot, but 4th order moment -> too sensitive to noise
- barkbands_skewness -> changes a lot, but 3th order moment -> too sensitive to noise
- barkbands_spread -> changes by ~1%, quite stable
- spectral_crest -> changes a lot -> uses the max which is way too sensitive to noise
- spectral_flatness_db -> changes a little (~5%)
- average loudness: very small deviation (<1%)
due to the framecutter not returning exactly the same number of frames (off by 1)
this difference does not appear on all the files (some have the exact same value)
- tempotap (and dependencies) has negligible variations (<0.01% for all dependencies)
- beatsloudness_bass varies a lot because we used energybandratio rounds to nearest
bin instead of truncating, hence we're not taking into account the bin nº 0,
which has lots of energy
Essentia 1.0.5:
- did a white-box review of 33 more algorithms
Essentia 1.0.5.1:
- Fixed the build on various platforms.
Essentia 1.0.6:
- did a white-box review of the remaining algorithms
----------------------------------------------------------------------
Essentia 1.1:
- Internal refactoring: Parameter, Pool, python bindings are all much cleaner now
- New algorithms: Panning, PCA, SpectralContrast, RhythmTransform, YamlInput, SVMClassifier, PoolAggregator, new FFMpeg-based audioloader & writer
- Streaming Python bindings (ESX is dead)
- Python bindings gained online docstrings
- Win32 & Mac OS X packages
- bug fixes, more (& better) unittests, etc...
----------------------------------------------------------------------
Essentia 1.2:
- Much better Windows & MacOSX installers!
- Better compatibility with python 2.6
- Core library changes:
- pool.merge (between pools and descriptors), pool.removeNamespace. See src/base/pool.*
- PoolAggregator also allows Kurtosis and Skewness as aggregates.
- audio loaders can now directly load the audio track from a video (works with youtube videos, etc...)
- new GaiaTransform algorithm that can use any classification model trained by Gaia to classify a given Pool. See src/algorithms/highlevel/gaiatransform.*
- spline (b, beta, quadratic) and cubicspline approximants. See src/algorithms/standard/spline and src/algorithms/standard/cubicspline
- streaming_extractor sub-extractors (lowlevelspectralextractor, lowleveleqlouspectralextractor, rhtythmdescriptors, tuningfrequencyextractor, tonalextractor, etc.) available as highlevel algorithms. See src/algorithms/highlevel
- External features / examples enhancements:
- Most of Essentia descriptors are now wrapped as vamp plugins. see src/examples/vamp*
- python to c++ translation and dot graphs (ala esx, i.e. only for composite algorithms not for actual python code). See src/examples/python for more details.
- streaming_extractor completely available in python (exact same results as its c++ counterparts). See src/examples/python/streaming_extractor for more details.
----------------------------------------------------------------------
Essentia 1.3:
- New rhythm descriptors, bpm and beat tracking using Meinhard Mueller's technique described in his paper from ISMIR 2009.
- streaming_extractor can now be configured using a YAML configuration file
- the usual assortment of bugfixes.
----------------------------------------------------------------------
Essentia 2.0:
- First release to be publicly available as free software released under AGPLv3.
- Refactoring of the core API
- fix small API annoyances for the standard mode
- streaming mode refactor. It is now much better defined, using sound computer science techniques (The visible
network is a directed acyclic graph, the composites have better defined semantics, and the order of execution of the
algorithms is the topological sort of the transitive reduction of the visible network after the composites have been
expanded). In particular, the scheduler that runs the algorithms in the streaming mode is now a lot more correct,
which permitted to clean all the small hacks that had accumulated in the algorithms themselves during the 1.x
releases to compensate for the deficiencies of the initial scheduler.
- New algorithms for onset detection, beat tracking and melody extraction
- New and updated features extractors
- Updated Vamp plugin
- Much better documentation, more python examples
- Bugfixes, more unittests, etc.