Accuracies and confusion matrices for default models

Read more about the model training process on the essentia website.

danceability

In-house MTG collection

Use: classification of music by danceability

Size: 306 full tracks, 124/182 per class

Accuracy: 92.410714%

Predicted (%)

danceable not_danceable Proportion
danceable 92.31 132 danceable (out of 143) classified as danceable 7.69 11 danceable (out of 143) classified as not_danceable danceable 63.84 %
not_danceable 7.41 6 not_danceable (out of 81) classified as danceable 92.59 75 not_danceable (out of 81) classified as not_danceable not_danceable36.16 %

Actual (%)


gender

In-house MTG collection

Use: classification of vocal music by gender (male/female)

Size: 3311 full tracks, 1508/1803 per class

Accuracy: 87.213915%

Predicted (%)

female male Proportion
female 91.20 1627 female (out of 1784) classified as female 8.80 157 female (out of 1784) classified as male female 54.44 %
male 17.55 262 male (out of 1493) classified as female 82.45 1231 male (out of 1493) classified as male male 45.56 %

Actual (%)


genre_dortmund

Music Audio Benchmark Data Set by TU Dortmund University (Homburg et al., 2005)

Use: classification of music by genre

Size: 1820 track excerpts, 46-490 per genre

Homburg, H., Mierswa, I., Möller, B., Morik, K., & Wurst, M. (2005). A Benchmark Dataset for Audio Classification and Clustering. In 6th International Conference on Music Information Retrieval (ISMIR'05), pp. 528-31.

Accuracy: 60.254372%

Predicted (%)

alternative blues electronic folkcountry funksoulrnb jazz pop raphiphop rock Proportion
alternative 10.27 15 alternative (out of 146) classified as alternative 4.79 7 alternative (out of 146) classified as blues 4.79 7 alternative (out of 146) classified as electronic 17.81 26 alternative (out of 146) classified as folkcountry 0.00 0 alternative (out of 146) classified as funksoulrnb 10.96 16 alternative (out of 146) classified as jazz 2.05 3 alternative (out of 146) classified as pop 1.37 2 alternative (out of 146) classified as raphiphop 47.95 70 alternative (out of 146) classified as rock alternative 7.74 %
blues 1.67 2 blues (out of 120) classified as alternative 35.83 43 blues (out of 120) classified as blues 0.83 1 blues (out of 120) classified as electronic 10.00 12 blues (out of 120) classified as folkcountry 0.83 1 blues (out of 120) classified as funksoulrnb 19.17 23 blues (out of 120) classified as jazz 0.00 0 blues (out of 120) classified as pop 2.50 3 blues (out of 120) classified as raphiphop 29.17 35 blues (out of 120) classified as rock blues 6.36 %
electronic 9.73 11 electronic (out of 113) classified as alternative 0.00 0 electronic (out of 113) classified as blues 39.82 45 electronic (out of 113) classified as electronic 5.31 6 electronic (out of 113) classified as folkcountry 0.00 0 electronic (out of 113) classified as funksoulrnb 17.70 20 electronic (out of 113) classified as jazz 0.88 1 electronic (out of 113) classified as pop 12.39 14 electronic (out of 113) classified as raphiphop 14.16 16 electronic (out of 113) classified as rock electronic 5.99 %
folkcountry 4.50 10 folkcountry (out of 222) classified as alternative 4.95 11 folkcountry (out of 222) classified as blues 0.45 1 folkcountry (out of 222) classified as electronic 56.76 126 folkcountry (out of 222) classified as folkcountry 0.00 0 folkcountry (out of 222) classified as funksoulrnb 12.16 27 folkcountry (out of 222) classified as jazz 3.15 7 folkcountry (out of 222) classified as pop 1.35 3 folkcountry (out of 222) classified as raphiphop 16.67 37 folkcountry (out of 222) classified as rock folkcountry 11.76 %
funksoulrnb 0.00 0 funksoulrnb (out of 47) classified as alternative 2.13 1 funksoulrnb (out of 47) classified as blues 2.13 1 funksoulrnb (out of 47) classified as electronic 4.26 2 funksoulrnb (out of 47) classified as folkcountry 2.13 1 funksoulrnb (out of 47) classified as funksoulrnb 31.91 15 funksoulrnb (out of 47) classified as jazz 2.13 1 funksoulrnb (out of 47) classified as pop 14.89 7 funksoulrnb (out of 47) classified as raphiphop 40.43 19 funksoulrnb (out of 47) classified as rock funksoulrnb 2.49 %
jazz 3.13 10 jazz (out of 319) classified as alternative 3.45 11 jazz (out of 319) classified as blues 4.08 13 jazz (out of 319) classified as electronic 8.78 28 jazz (out of 319) classified as folkcountry 0.63 2 jazz (out of 319) classified as funksoulrnb 67.08 214 jazz (out of 319) classified as jazz 1.57 5 jazz (out of 319) classified as pop 3.45 11 jazz (out of 319) classified as raphiphop 7.84 25 jazz (out of 319) classified as rock jazz 16.91 %
pop 6.03 7 pop (out of 116) classified as alternative 4.31 5 pop (out of 116) classified as blues 3.45 4 pop (out of 116) classified as electronic 28.45 33 pop (out of 116) classified as folkcountry 0.00 0 pop (out of 116) classified as funksoulrnb 4.31 5 pop (out of 116) classified as jazz 8.62 10 pop (out of 116) classified as pop 8.62 10 pop (out of 116) classified as raphiphop 36.21 42 pop (out of 116) classified as rock pop 6.15 %
raphiphop 0.33 1 raphiphop (out of 300) classified as alternative 1.33 4 raphiphop (out of 300) classified as blues 2.33 7 raphiphop (out of 300) classified as electronic 0.67 2 raphiphop (out of 300) classified as folkcountry 0.00 0 raphiphop (out of 300) classified as funksoulrnb 1.67 5 raphiphop (out of 300) classified as jazz 0.67 2 raphiphop (out of 300) classified as pop 88.00 264 raphiphop (out of 300) classified as raphiphop 5.00 15 raphiphop (out of 300) classified as rock raphiphop 15.90 %
rock 3.37 17 rock (out of 504) classified as alternative 2.58 13 rock (out of 504) classified as blues 1.19 6 rock (out of 504) classified as electronic 4.76 24 rock (out of 504) classified as folkcountry 0.00 0 rock (out of 504) classified as funksoulrnb 3.37 17 rock (out of 504) classified as jazz 0.60 3 rock (out of 504) classified as pop 0.99 5 rock (out of 504) classified as raphiphop 83.13 419 rock (out of 504) classified as rock rock 26.71 %

Actual (%)


genre_electronic

In-house MTG collection

Use: classification of electronic music by subgenres

Size: 250 track excerpts, 50 per genre

Accuracy: 91.699605%

Predicted (%)

ambient dnb house techno trance Proportion
ambient 98.11 52 ambient (out of 53) classified as ambient 0.00 0 ambient (out of 53) classified as dnb 0.00 0 ambient (out of 53) classified as house 1.89 1 ambient (out of 53) classified as techno 0.00 0 ambient (out of 53) classified as trance ambient 20.95 %
dnb 0.00 0 dnb (out of 50) classified as ambient 96.00 48 dnb (out of 50) classified as dnb 0.00 0 dnb (out of 50) classified as house 2.00 1 dnb (out of 50) classified as techno 2.00 1 dnb (out of 50) classified as trance dnb 19.76 %
house 0.00 0 house (out of 50) classified as ambient 2.00 1 house (out of 50) classified as dnb 94.00 47 house (out of 50) classified as house 4.00 2 house (out of 50) classified as techno 0.00 0 house (out of 50) classified as trance house 19.76 %
techno 0.00 0 techno (out of 50) classified as ambient 8.00 4 techno (out of 50) classified as dnb 12.00 6 techno (out of 50) classified as house 80.00 40 techno (out of 50) classified as techno 0.00 0 techno (out of 50) classified as trance techno 19.76 %
trance 0.00 0 trance (out of 50) classified as ambient 4.00 2 trance (out of 50) classified as dnb 4.00 2 trance (out of 50) classified as house 2.00 1 trance (out of 50) classified as techno 90.00 45 trance (out of 50) classified as trance trance 19.76 %

Actual (%)


genre_rosamerica

In-house MTG collection created by a musicologist (Guaus, 2009)

Use: classification of music by genre

Size: 400 tracks, 50 per genre

Classes: classical, dance, hip-hop, jazz, pop, rhythm'n'blues, rock, speech

Guaus, E. (2009). Audio content processing for automatic music genre classification: descriptors, databases, and classifiers (Doctoral dissertation, Universitat Pompeu Fabra, Barcelona).

Accuracy: 87.557604%

Predicted (%)

cla dan hip jaz pop rhy roc spe Proportion
cla 92.73 51 cla (out of 55) classified as cla 0.00 0 cla (out of 55) classified as dan 0.00 0 cla (out of 55) classified as hip 1.82 1 cla (out of 55) classified as jaz 1.82 1 cla (out of 55) classified as pop 1.82 1 cla (out of 55) classified as rhy 1.82 1 cla (out of 55) classified as roc 0.00 0 cla (out of 55) classified as spe cla 12.67 %
dan 0.00 0 dan (out of 54) classified as cla 96.30 52 dan (out of 54) classified as dan 1.85 1 dan (out of 54) classified as hip 0.00 0 dan (out of 54) classified as jaz 0.00 0 dan (out of 54) classified as pop 1.85 1 dan (out of 54) classified as rhy 0.00 0 dan (out of 54) classified as roc 0.00 0 dan (out of 54) classified as spe dan 12.44 %
hip 0.00 0 hip (out of 55) classified as cla 1.82 1 hip (out of 55) classified as dan 94.55 52 hip (out of 55) classified as hip 0.00 0 hip (out of 55) classified as jaz 0.00 0 hip (out of 55) classified as pop 3.64 2 hip (out of 55) classified as rhy 0.00 0 hip (out of 55) classified as roc 0.00 0 hip (out of 55) classified as spe hip 12.67 %
jaz 3.77 2 jaz (out of 53) classified as cla 0.00 0 jaz (out of 53) classified as dan 0.00 0 jaz (out of 53) classified as hip 90.57 48 jaz (out of 53) classified as jaz 5.66 3 jaz (out of 53) classified as pop 0.00 0 jaz (out of 53) classified as rhy 0.00 0 jaz (out of 53) classified as roc 0.00 0 jaz (out of 53) classified as spe jaz 12.21 %
pop 1.82 1 pop (out of 55) classified as cla 7.27 4 pop (out of 55) classified as dan 1.82 1 pop (out of 55) classified as hip 0.00 0 pop (out of 55) classified as jaz 69.09 38 pop (out of 55) classified as pop 16.36 9 pop (out of 55) classified as rhy 3.64 2 pop (out of 55) classified as roc 0.00 0 pop (out of 55) classified as spe pop 12.67 %
rhy 1.85 1 rhy (out of 54) classified as cla 3.70 2 rhy (out of 54) classified as dan 5.56 3 rhy (out of 54) classified as hip 11.11 6 rhy (out of 54) classified as jaz 11.11 6 rhy (out of 54) classified as pop 66.67 36 rhy (out of 54) classified as rhy 0.00 0 rhy (out of 54) classified as roc 0.00 0 rhy (out of 54) classified as spe rhy 12.44 %
roc 0.00 0 roc (out of 54) classified as cla 1.85 1 roc (out of 54) classified as dan 0.00 0 roc (out of 54) classified as hip 0.00 0 roc (out of 54) classified as jaz 7.41 4 roc (out of 54) classified as pop 0.00 0 roc (out of 54) classified as rhy 90.74 49 roc (out of 54) classified as roc 0.00 0 roc (out of 54) classified as spe roc 12.44 %
spe 0.00 0 spe (out of 54) classified as cla 0.00 0 spe (out of 54) classified as dan 0.00 0 spe (out of 54) classified as hip 0.00 0 spe (out of 54) classified as jaz 0.00 0 spe (out of 54) classified as pop 0.00 0 spe (out of 54) classified as rhy 0.00 0 spe (out of 54) classified as roc 100.00 54 spe (out of 54) classified as spe spe 12.44 %

Actual (%)


genre_tzanetakis

GTZAN Genre Collection (Tzanetakis and Cook, 2002)

Use: classification of music by genre

Size: 1000 track excerpts, 100 per genre

Tzanetakis, G., & Cook, P. (2002). Musical genre classification of audio signals. IEEE transactions on Speech and Audio Processing, 10(5), 293-302.

Sturm, B. L. (2012). An analysis of the GTZAN music genre dataset. In 2nd International ACM Workshop on Music Information Retrieval with User-centered and Multimodal Strategies (pp. 7-12).

Accuracy: 75.528701%

Predicted (%)

< th>pop
blu cla cou dis hip jaz met pop reg roc Proportion
blu 78.00 78 blu (out of 100) classified as blu 1.00 1 blu (out of 100) classified as cla 8.00 8 blu (out of 100) classified as cou 3.00 3 blu (out of 100) classified as dis 1.00 1 blu (out of 100) classified as hip 1.00 1 blu (out of 100) classified as jaz 3.00 3 blu (out of 100) classified as met 0.00 0 blu (out of 100) classified as pop 2.00 2 blu (out of 100) classified as reg 3.00 3 blu (out of 100) classified as roc blu 10.07 %
cla 2.15 2 cla (out of 93) classified as blu 92.47 86 cla (out of 93) classified as cla 1.08 1 cla (out of 93) classified as cou 2.15 2 cla (out of 93) classified as dis 0.00 0 cla (out of 93) classified as hip 0.00 0 cla (out of 93) classified as jaz 0.00 0 cla (out of 93) classified as met 1.08 1 cla (out of 93) classified as pop 0.00 0 cla (out of 93) classified as reg 1.08 1 cla (out of 93) classified as roc cla 9.37 %
cou 1.00 1 cou (out of 100) classified as blu 1.00 1 cou (out of 100) classified as cla 78.00 78 cou (out of 100) classified as cou 7.00 7 cou (out of 100) classified as dis 0.00 0 cou (out of 100) classified as hip 2.00 2 cou (out of 100) classified as jaz 0.00 0 cou (out of 100) classified as met 4.00 4 cou (out of 100) classified as pop 3.00 3 cou (out of 100) classified as reg 4.00 4 cou (out of 100) classified as roc cou 10.07 %
dis 0.00 0 dis (out of 100) classified as blu 1.00 1 dis (out of 100) classified as cla 5.00 5 dis (out of 100) classified as cou 71.00 71 dis (out of 100) classified as dis 3.00 3 dis (out of 100) classified as hip 1.00 1 dis (out of 100) classified as jaz 2.00 2 dis (out of 100) classified as met 7.00 7 dis (out of 100) classified as pop 5.00 5 dis (out of 100) classified as reg 5.00 5 dis (out of 100) classified as roc dis 10.07 %
hip 2.00 2 hip (out of 100) classified as blu 1.00 1 hip (out of 100) classified as cla 0.00 0 hip (out of 100) classified as cou 6.00 6 hip (out of 100) classified as dis 73.00 73 hip (out of 100) classified as hip 0.00 0 hip (out of 100) classified as jaz 3.00 3 hip (out of 100) classified as met 3.00 3 hip (out of 100) classified as pop 11.00 11 hip (out of 100) classified as reg 1.00 1 hip (out of 100) classified as roc hip 10.07 %
jaz 7.00 7 jaz (out of 100) classified as blu 4.00 4 jaz (out of 100) classified as cla 4.00 4 jaz (out of 100) classified as cou 3.00 3 jaz (out of 100) classified as dis 1.00 1 jaz (out of 100) classified as hip 79.00 79 jaz (out of 100) classified as jaz 0.00 0 jaz (out of 100) classified as met 1.00 1 jaz (out of 100) classified as pop 1.00 1 jaz (out of 100) classified as reg 0.00 0 jaz (out of 100) classified as roc jaz 10.07 %
met 2.00 2 met (out of 100) classified as blu 0.00 0 met (out of 100) classified as cla 0.00 0 met (out of 100) classified as cou 1.00 1 met (out of 100) classified as dis 3.00 3 met (out of 100) classified as hip 2.00 2 met (out of 100) classified as jaz 86.00 86 met (out of 100) classified as met 1.00 1 met (out of 100) classified as pop 0.00 0 met (out of 100) classified as reg 5.00 5 met (out of 100) classified as roc met 10.07 %
pop 0.00 0 pop (out of 100) classified as blu 1.00 1 pop (out of 100) classified as cla 6.00 6 pop (out of 100) classified as cou 6.00 6 pop (out of 100) classified as dis 5.00 5 pop (out of 100) classified as hip 0.00 0 pop (out of 100) classified as jaz 0.00 0 pop (out of 100) classified as met 75.00 75 pop (out of 100) classified as pop 4.00 4 pop (out of 100) classified as reg 3.00 3 pop (out of 100) classified as roc 10.07 %
reg 3.00 3 reg (out of 100) classified as blu 2.00 2 reg (out of 100) classified as cla 4.00 4 reg (out of 100) classified as cou 4.00 4 reg (out of 100) classified as dis 11.00 11 reg (out of 100) classified as hip 2.00 2 reg (out of 100) classified as jaz 0.00 0 reg (out of 100) classified as met 5.00 5 reg (out of 100) classified as pop 64.00 64 reg (out of 100) classified as reg 5.00 5 reg (out of 100) classified as roc reg 10.07 %
roc 7.00 7 roc (out of 100) classified as blu 2.00 2 roc (out of 100) classified as cla 6.00 6 roc (out of 100) classified as cou 10.00 10 roc (out of 100) classified as dis 3.00 3 roc (out of 100) classified as hip 2.00 2 roc (out of 100) classified as jaz 4.00 4 roc (out of 100) classified as met 2.00 2 roc (out of 100) classified as pop 4.00 4 roc (out of 100) classified as reg 60.00 60 roc (out of 100) classified as roc roc 10.07 %

Actual (%)


ismir04_rhythm

ISMIR2004 Rhythm Classification Dataset ("Ballroom dataset") (Cano et al., 2006)

Use: classification of ballroom music by dance styles

Size: 683 track excerpts, 60-110 per class

Cano, P., Gómez, E., Gouyon, F., Herrera, P., Koppenberger, M., Ong, B., ... & Wack, N. (2006). ISMIR 2004 audio description contest. Music Technology Group, Universitat Pompeu Fabra, Tech. Rep.

Accuracy: 73.209169%

Predicted (%)

ChaChaCha Jive Quickstep Rumba-American Rumba-International Rumba-Misc Samba Tango VienneseWaltz Waltz Proportion
ChaChaCha 83.78 93 ChaChaCha (out of 111) classified as ChaChaCha 5.41 6 ChaChaCha (out of 111) classified as Jive 1.80 2 ChaChaCha (out of 111) classified as Quickstep 0.00 0 ChaChaCha (out of 111) classified as Rumba-American 3.60 4 ChaChaCha (out of 111) classified as Rumba-International 0.00 0 ChaChaCha (out of 111) classified as Rumba-Misc 3.60 4 ChaChaCha (out of 111) classified as Samba 1.80 2 ChaChaCha (out of 111) classified as Tango 0.00 0 ChaChaCha (out of 111) classified as VienneseWaltz 0.00 0 ChaChaCha (out of 111) classified as Waltz ChaChaCha 15.90 %
Jive 15.00 9 Jive (out of 60) classified as ChaChaCha 68.33 41 Jive (out of 60) classified as Jive 3.33 2 Jive (out of 60) classified as Quickstep 0.00 0 Jive (out of 60) classified as Rumba-American 3.33 2 Jive (out of 60) classified as Rumba-International 1.67 1 Jive (out of 60) classified as Rumba-Misc 5.00 3 Jive (out of 60) classified as Samba 0.00 0 Jive (out of 60) classified as Tango 3.33 2 Jive (out of 60) classified as VienneseWaltz 0.00 0 Jive (out of 60) classified as Waltz Jive 8.60 %
Quickstep 6.10 5 Quickstep (out of 82) classified as ChaChaCha 3.66 3 Quickstep (out of 82) classified as Jive 74.39 61 Quickstep (out of 82) classified as Quickstep 1.22 1 Quickstep (out of 82) classified as Rumba-American 2.44 2 Quickstep (out of 82) classified as Rumba-International 0.00 0 Quickstep (out of 82) classified as Rumba-Misc 10.98 9 Quickstep (out of 82) classified as Samba 0.00 0 Quickstep (out of 82) classified as Tango 0.00 0 Quickstep (out of 82) classified as VienneseWaltz 1.22 1 Quickstep (out of 82) classified as Waltz Quickstep 11.75 %
Rumba-American 0.00 0 Rumba-American (out of 7) classified as ChaChaCha 0.00 0 Rumba-American (out of 7) classified as Jive 14.29 1 Rumba-American (out of 7) classified as Quickstep 28.57 2 Rumba-American (out of 7) classified as Rumba-American 42.86 3 Rumba-American (out of 7) classified as Rumba-International 0.00 0 Rumba-American (out of 7) classified as Rumba-Misc 0.00 0 Rumba-American (out of 7) classified as Samba 14.29 1 Rumba-American (out of 7) classified as Tango 0.00 0 Rumba-American (out of 7) classified as VienneseWaltz 0.00 0 Rumba-American (out of 7) classified as Waltz Rumba-American1.00 %
Rumba-International 1.96 1 Rumba-International (out of 51) classified as ChaChaCha 0.00 0 Rumba-International (out of 51) classified as Jive 3.92 2 Rumba-International (out of 51) classified as Quickstep 1.96 1 Rumba-International (out of 51) classified as Rumba-American 74.51 38 Rumba-International (out of 51) classified as Rumba-International 0.00 0 Rumba-International (out of 51) classified as Rumba-Misc 1.96 1 Rumba-International (out of 51) classified as Samba 1.96 1 Rumba-International (out of 51) classified as Tango 7.84 4 Rumba-International (out of 51) classified as VienneseWaltz 5.88 3 Rumba-International (out of 51) classified as Waltz Rumba-International 7.31 %
Rumba-Misc 5.00 2 Rumba-Misc (out of 40) classified as ChaChaCha 7.50 3 Rumba-Misc (out of 40) classified as Jive 7.50 3 Rumba-Misc (out of 40) classified as Quickstep 0.00 0 Rumba-Misc (out of 40) classified as Rumba-American 2.50 1 Rumba-Misc (out of 40) classified as Rumba-International 52.50 21 Rumba-Misc (out of 40) classified as Rumba-Misc 5.00 2 Rumba-Misc (out of 40) classified as Samba 5.00 2 Rumba-Misc (out of 40) classified as Tango 2.50 1 Rumba-Misc (out of 40) classified as VienneseWaltz 12.50 5 Rumba-Misc (out of 40) classified as Waltz Rumba-Misc 5.73 %
Samba 6.98 6 Samba (out of 86) classified as ChaChaCha 6.98 6 Samba (out of 86) classified as Jive 13.95 12 Samba (out of 86) classified as Quickstep 0.00 0 Samba (out of 86) classified as Rumba-American 2.33 2 Samba (out of 86) classified as Rumba-International 1.16 1 Samba (out of 86) classified as Rumba-Misc 66.28 57 Samba (out of 86) classified as Samba 2.33 2 Samba (out of 86) classified as Tango 0.00 0 Samba (out of 86) classified as VienneseWaltz 0.00 0 Samba (out of 86) classified as Waltz Samba 12.32 %
Tango 4.65 4 Tango (out of 86) classified as ChaChaCha 0.00 0 Tango (out of 86) classified as Jive 0.00 0 Tango (out of 86) classified as Quickstep 2.33 2 Tango (out of 86) classified as Rumba-American 0.00 0 Tango (out of 86) classified as Rumba-International 5.81 5 Tango (out of 86) classified as Rumba-Misc 0.00 0 Tango (out of 86) classified as Samba 83.72 72 Tango (out of 86) classified as Tango 1.16 1 Tango (out of 86) classified as VienneseWaltz 2.33 2 Tango (out of 86) classified as Waltz Tango 12.32 %
VienneseWaltz 0.00 0 VienneseWaltz (out of 65) classified as ChaChaCha 1.54 1 VienneseWaltz (out of 65) classified as Jive 1.54 1 VienneseWaltz (out of 65) classified as Quickstep 0.00 0 VienneseWaltz (out of 65) classified as Rumba-American 4.62 3 VienneseWaltz (out of 65) classified as Rumba-International 3.08 2 VienneseWaltz (out of 65) classified as Rumba-Misc 0.00 0 VienneseWaltz (out of 65) classified as Samba 0.00 0 VienneseWaltz (out of 65) classified as Tango 67.69 44 VienneseWaltz (out of 65) classified as VienneseWaltz 21.54 14 VienneseWaltz (out of 65) classified as Waltz VienneseWaltz 9.31 %
Waltz 0.00 0 Waltz (out of 110) classified as ChaChaCha 0.00 0 Waltz (out of 110) classified as Jive 0.91 1 Waltz (out of 110) classified as Quickstep 0.00 0 Waltz (out of 110) classified as Rumba-American 4.55 5 Waltz (out of 110) classified as Rumba-International 6.36 7 Waltz (out of 110) classified as Rumba-Misc 0.00 0 Waltz (out of 110) classified as Samba 2.73 3 Waltz (out of 110) classified as Tango 10.91 12 Waltz (out of 110) classified as VienneseWaltz 74.55 82 Waltz (out of 110) classified as Waltz Waltz 15.76 %

Actual (%)


mood_acoustic

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by type of sound (acoustic/non-acoustic)

Size: 321 full tracks + excerpts, 193/128 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 92.982456%

Predicted (%)

acoustic not_acoustic Proportion
acoustic 95.04 115 acoustic (out of 121) classified as acoustic 4.96 6 acoustic (out of 121) classified as not_acoustic acoustic 53.07 %
not_acoustic 9.35 10 not_acoustic (out of 107) classified as acoustic 90.65 97 not_acoustic (out of 107) classified as not_acoustic not_acoustic46.93 %

Actual (%)


mood_aggressive

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by mood (aggressive/non-aggressive)

Size: 280 full tracks + excerpts, 133/147 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 97.500000%

Predicted (%)

aggressive not_aggressive Proportion
aggressive 95.49 127 aggressive (out of 133) classified as aggressive 4.51 6 aggressive (out of 133) classified as not_aggressive aggressive 47.50 %
not_aggressive 0.68 1 not_aggressive (out of 147) classified as aggressive 99.32 146 not_aggressive (out of 147) classified as not_aggressive not_aggressive 52.50 %

Actual (%)


mood_electronic

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by type of sound (electronic/non-electronic)

Size: 332 full tracks + excerpts, 164/168 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 86.379928%

Predicted (%)

electronic not_electronic Proportion
electronic 82.84 111 electronic (out of 134) classified as electronic 17.16 23 electronic (out of 134) classified as not_electronic electronic 48.03 %
not_electronic 10.34 15 not_electronic (out of 145) classified as electronic 89.66 130 not_electronic (out of 145) classified as not_electronic not_electronic 51.97 %

Actual (%)


mood_happy

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by mood (happy/non-happy)

Size: 302 full tracks + excerpts, 139/163 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 83.265306%

Predicted (%)

happy not_happy Proportion
happy 82.14 92 happy (out of 112) classified as happy 17.86 20 happy (out of 112) classified as not_happy happy 45.71 %
not_happy 15.79 21 not_happy (out of 133) classified as happy 84.21 112 not_happy (out of 133) classified as not_happy not_happy54.29 %

Actual (%)


mood_party

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by mood (party/non-party)

Size: 349 full tracks + excerpts, 198/151 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 88.381743%

Predicted (%)

not_party party Proportion
not_party 85.07 114 not_party (out of 134) classified as not_party 14.93 20 not_party (out of 134) classified as party not_party 55.60 %
party 7.48 8 party (out of 107) classified as not_party 92.52 99 party (out of 107) classified as party party44.40 %

Actual (%)


mood_relaxed

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by mood (relaxed/non-relaxed)

Size: 446 full tracks + excerpts, 145/301 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 93.201133%

Predicted (%)

not_relaxed relaxed Proportion
not_relaxed 85.48 106 not_relaxed (out of 124) classified as not_relaxed 14.52 18 not_relaxed (out of 124) classified as relaxed not_relaxed 35.13 %
relaxed 2.62 6 relaxed (out of 229) classified as not_relaxed 97.38 223 relaxed (out of 229) classified as relaxed relaxed 64.87 %

Actual (%)


mood_sad

In-house MTG collection (Laurier et al., 2009)

Use: classification of music by mood (sad/non-sad)

Size: 230 full tracks + excerpts, 96/134 per class

Laurier, C., Meyers, O., Serra, J., Blech, M., & Herrera, P. (2009). Music mood annotator design and integration. In 7th International Workshop on Content-Based Multimedia Indexing (CBMI'09), pp. 156-161.

Accuracy: 87.826087%

Predicted (%)

not_sad sad Proportion
not_sad 82.29 79 not_sad (out of 96) classified as not_sad 17.71 17 not_sad (out of 96) classified as sad not_sad 41.74 %
sad 8.21 11 sad (out of 134) classified as not_sad 91.79 123 sad (out of 134) classified as sad sad 58.26 %

Actual (%)


moods_mirex

MIREX Audio Mood Classification Dataset (Hu and Downie, 2007)

Use: classification of music into 5 mood clusters

Size: 269 track excerpts, 60-110 per class

Hu, X., & Downie, J. S. (2007). Exploring Mood Metadata: Relationships with Genre, Artist and Usage Metadata. In 8th International Conference on Music Information Retrieval (ISMIR'07), pp. 67-72.

Accuracy: 57.089552%

Predicted (%)

Cluster1 Cluster2 Cluster3 Cluster4 Cluster5 Proportion
Cluster1 56.90 33 Cluster1 (out of 58) classified as Cluster1 15.52 9 Cluster1 (out of 58) classified as Cluster2 8.62 5 Cluster1 (out of 58) classified as Cluster3 5.17 3 Cluster1 (out of 58) classified as Cluster4 13.79 8 Cluster1 (out of 58) classified as Cluster5 Cluster1 21.64 %
Cluster2 22.22 12 Cluster2 (out of 54) classified as Cluster1 53.70 29 Cluster2 (out of 54) classified as Cluster2 16.67 9 Cluster2 (out of 54) classified as Cluster3 5.56 3 Cluster2 (out of 54) classified as Cluster4 1.85 1 Cluster2 (out of 54) classified as Cluster5 Cluster2 20.15 %
Cluster3 5.48 4 Cluster3 (out of 73) classified as Cluster1 19.18 14 Cluster3 (out of 73) classified as Cluster2 68.49 50 Cluster3 (out of 73) classified as Cluster3 5.48 4 Cluster3 (out of 73) classified as Cluster4 1.37 1 Cluster3 (out of 73) classified as Cluster5 Cluster3 27.24 %
Cluster4 15.62 5 Cluster4 (out of 32) classified as Cluster1 37.50 12 Cluster4 (out of 32) classified as Cluster2 15.62 5 Cluster4 (out of 32) classified as Cluster3 25.00 8 Cluster4 (out of 32) classified as Cluster4 6.25 2 Cluster4 (out of 32) classified as Cluster5 Cluster4 11.94 %
Cluster5 27.45 14 Cluster5 (out of 51) classified as Cluster1 7.84 4 Cluster5 (out of 51) classified as Cluster2 0.00 0 Cluster5 (out of 51) classified as Cluster3 0.00 0 Cluster5 (out of 51) classified as Cluster4 64.71 33 Cluster5 (out of 51) classified as Cluster5 Cluster519.03 %

Actual (%)


timbre

In-house MTG collection

Use: classification of music by timbre colour (dark/bright timbre)

Size: 3000 track excerpts, 1500 per class

Accuracy: 94.317418%

Predicted (%)

bright dark Proportion
bright 93.76 1383 bright (out of 1475) classified as bright 6.24 92 bright (out of 1475) classified as dark bright 49.60 %
dark 5.14 77 dark (out of 1499) classified as bright 94.86 1422 dark (out of 1499) classified as dark dark50.40 %

Actual (%)


tonal_atonal

In-house MTG collection

Use: classification of music by tonality (tonal/atonal)

Size: 345 track excerpts, 200/145

Accuracy: 97.667638%

Predicted (%)

atonal tonal Proportion
atonal 96.53 139 atonal (out of 144) classified as atonal 3.47 5 atonal (out of 144) classified as tonal atonal 41.98 %
tonal 1.51 3 tonal (out of 199) classified as atonal 98.49 196 tonal (out of 199) classified as tonal tonal 58.02 %

Actual (%)


voice_instrumental

In-house MTG collection

Use: classification into music with voice/instrumental

Size: 1000 track excerpts, 500 per class

Accuracy: 93.800000%

Predicted (%)

instrumental voice Proportion
instrumental 94.20 471 instrumental (out of 500) classified as instrumental 5.80 29 instrumental (out of 500) classified as voice instrumental 50.00 %
voice 6.60 33 voice (out of 500) classified as instrumental 93.40 467 voice (out of 500) classified as voice voice50.00 %

Actual (%)