Difference between revisions of "2009:Audio Cover Song Identification Results"

From MIREX Wiki
(→‎Individual Results Files)
m (Robot: Automated text replacement (-mirex/abs/ +mirex/abstracts/))
 
(62 intermediate revisions by 5 users not shown)
Line 1: Line 1:
 +
== Introduction ==
  
−
== Introduction ==
+
This year, we ran Audio Cover Song (ACS) Identfication with two datasets: ''Mixed Collection'' and ''Sapp's Mazurka Collection''.
 +
 
 +
===Mixed Collection Information===
 +
This is the "original" ACS collection. Within the 1000 pieces in the Audio Cover Song database, there are embedded 30 different "cover songs" each represented by 11 different "versions" for a total of 330 audio files (16bit, monophonic, 22.05khz, wav). The "cover songs" represent a variety of genres (e.g., classical, jazz, gospel, rock, folk-rock, etc.) and the variations span a variety of styles and orchestrations.
 +
 
 +
Using each of these cover song files in turn as as the "seed/query" file, we will examine the returned lists of items for the presence of the other 10 versions of the "seed/query" file.
 +
 
 +
===Sapp's Mazurka Collection Information ===
 +
In addition to our original ACS dataset, we used the  [http://www.mazurka.org.uk/ Mazurka.org dataset] put together by Craig Sapp. We randomly chose 11 versions from 49 mazurkas and ran it as a separate ACS subtask. Systems returned a distance matrix of 539x539 from which we located the ranks of each of the associated cover versions.
  
 
=== General Legend ===
 
=== General Legend ===
Line 6: Line 15:
 
====Team ID====
 
====Team ID====
  
−
'''RE''' [[Suman Ravuri]]
+
'''RE''' = [https://www.music-ir.org/mirex/abstracts/2009/RE.pdf Suman Ravuri, Daniel P. W. Ellis] <br />
 +
'''SZA''' = [https://www.music-ir.org/mirex/abstracts/2009/SZA.pdf Joan Serra, Massimiliano Zanin, Ralph G. Andrzejak]<br />
 +
'''TA''' = [https://www.music-ir.org/mirex/abstracts/2009/ACS_ahonen.pdf Teppo E. Ahonen]<br />
 +
 
 +
== Results ==
 +
 
 +
===Mixed Collection===
 +
 
 +
====Summary Results====
 +
<csv p=2>2009//cover/coversong.mixed.summary.csv</csv>
 +
 
 +
====Number of Correct Covers at Rank X Returned in Top Ten====
 +
<csv>2009/cover/coversong.mixed.toptendist.csv</csv>
 +
 
 +
====Average Performance per Query Group====
 +
<csv p=2>2009/cover/coversong.mixed.precision.groups.csv</csv>
  
−
'''SZA''' [[Joan Serr`a1,2, Massimiliano Zanin3, and Ralph G. Andrzejak2]]
 
  
−
'''TA''' [[Teppo E. Ahonen]]
+
====Friedman's Test for Significant Differences====
 +
The Friedman test was run in MATLAB against the Average Precision summary data over the 30 song groups.<br /> Command: [c,m,h,gnames] = multcompare(stats, 'ctype', 'tukey-kramer','estimate', 'friedman', 'alpha', 0.05);
  
−
== Results ==
+
<csv p=2>2009//covertests/Coversong_Mixed_Ave_Performance_Per_Query_Group_friedman_tukeyKramerHSD.csv</csv>
  
−
===Mixed===
 
  
−
'''Summary results'''
+
[[Image:2009_coversong_mixed_ave_performance_per_query_group_friedman_mean_ranks.png]]
−
<csv p=2>/cover/coversong.mixed.summary.csv</csv>
 
  
−
'''Number of Correct Covers at Rank X Returned in Top Ten'''
+
====Run Times====
−
<csv>cover/coversong.toptendist.mixed.csv</csv>
+
TBA
  
−
===Mazurakas===
+
===Sapp's Mazuraka Collection===
  
−
'''Summary results'''
+
====Summary results====
−
<csv p=2>/cover/coversong.mazurka.summary.csv</csv>
+
<csv p=2>2009//cover/coversong.mazurka.summary.csv</csv>
  
−
'''Number of Correct Covers at Rank X Returned in Top Ten'''
+
====Number of Correct Covers at Rank X Returned in Top Ten====
−
<csv>cover/coversong.toptendist.mazurka.csv</csv>
+
<csv>2009/cover/coversong.mazurka.toptendist.csv</csv>
  
−
===Run Times===  
+
====Average Performance per Query Group====
 +
<csv p=2>2009/cover/coversong.mazurka.precision.groups.csv</csv>
  
  
−
===Friedman's Test for Significant Differences===
+
====Friedman's Test for Significant Differences====
 
The Friedman test was run in MATLAB against the Average Precision summary data over the 30 song groups.<br /> Command: [c,m,h,gnames] = multcompare(stats, 'ctype', 'tukey-kramer','estimate', 'friedman', 'alpha', 0.05);
 
The Friedman test was run in MATLAB against the Average Precision summary data over the 30 song groups.<br /> Command: [c,m,h,gnames] = multcompare(stats, 'ctype', 'tukey-kramer','estimate', 'friedman', 'alpha', 0.05);
  
−
<csv>cover/coversong.friedman.anova.csv</csv>
+
<csv>2009/covertests/Coversong_Mazurkas_Ave_Performance_Per_Query_Group_friedman_tukeyKramerHSD.csv</csv>
 +
 
 +
 
 +
[[Image:2009_coversong_mazurkas_ave_performance_per_query_group_friedman_mean_ranks.png]]
 +
 
 +
====Run Times====
 +
TBA
  
−
<csv>cover/coversong.friedman.csv</csv>
+
==Individual Results Files==
  
−
[[Image:coversong.friedman.png]]
+
===Mixed Collection===
 +
'''Average Precision by Query'''
  
−
===Average Performance per Query Group===
+
'''RE''' : [https://music-ir.org/mirex/results/2009/cover/RE.fix.mix.avgprec.txt S. Ravuri, D.P.W Ellis] <br />
−
These are the arithmetic means of the average precisions within each of the 30 query groups.
+
'''SZA''' : [https://music-ir.org/mirex/results/2009/cover/SZA.mix.avgprec.txt J. Serra et al.]<br />
 +
'''TA''' : [https://music-ir.org/mirex/results/2009/cover/TA.mix.avgprec.txt T. Ahonen] <br />
  
−
<csv>cover/cover.mapquerygroup.v2.csv</csv>
+
'''Average Precision by Cover Group'''
  
−
==Individual Results Files==
+
'''RE''' : [https://music-ir.org/mirex/results/2009/cover/RE.fix.mix.avgprec.groups.txt S. Ravuri, D.P.W Ellis]<br />
 +
'''SZA''' : [https://music-ir.org/mirex/results/2009/cover/SZA.mix.avgprec.groups.txt J. Serra et al.]<br />
 +
'''TA''' : [https://music-ir.org/mirex/results/2009/cover/TA.mix.avgprec.groups.txt T. Ahonen]<br />
 +
 
 +
'''Rank Lists'''
 +
 
 +
'''RE''' : [https://music-ir.org/mirex/results/2009/cover/RE.fix.mix.ranklist.txt S. Ravuri, D.P.W Ellis] <br />
 +
'''SZA''' : [https://music-ir.org/mirex/results/2009/cover/SZA.mix.ranklist.txt J. Serra et al.] <br />
 +
'''TA''' : [https://music-ir.org/mirex/results/2009/cover/TA.mix.ranklist.txt T. Ahonen]<br />
  
−
[/results/cover/RE.mix.avgprec.txt]
+
===Sapp's Mazurka Collection===
−
[/results/cover/RE.mix.ranklist.txt]
 
  
−
[/results/cover/SZA.mix.avgprec.txt]
+
'''Average Precision by Query'''
−
[/results/cover/SZA.mix.ranklist.txt]
 
  
−
[/results/cover/TA.mix.avgprec.txt]
+
'''RE''' : [https://music-ir.org/mirex/results/2009/cover/RE.maz.avgprec.txt S. Ravuri]<br />
−
[/results/cover/TA.mix.ranklist.txt]
+
'''SZA''' : [https://music-ir.org/mirex/results/2009/cover/SZA.maz.avgprec.txt J. Serra et al.]<br />
 +
'''TA''' : [https://music-ir.org/mirex/results/2009/cover/TA.maz.avgprec.txt T. Ahonen]<br />
  
 +
'''Average Precision by Cover Group'''
  
−
[/results/cover/RE.maz.avgprec.txt]
+
'''RE''' : [https://music-ir.org/mirex/results/2009/cover/RE.maz.avgprec.groups.txt S. Ravuri]<br />
−
[/results/cover/RE.maz.ranklist.txt]
+
'''SZA''' : [https://music-ir.org/mirex/results/2009/cover/SZA.maz.avgprec.groups.txt J. Serra et al.]<br />
 +
'''TA''' : [https://music-ir.org/mirex/results/2009/cover/TA.maz.avgprec.groups.txt T. Ahonen]<br />
  
−
[/results/cover/SZA.maz.avgprec.txt]
+
'''Rank Lists'''
−
[/results/cover/SZA.maz.ranklist.txt]
 
  
−
[/results/cover/TA.maz.avgprec.txt]
+
'''RE''' : [https://music-ir.org/mirex/results/2009/cover/RE.maz.ranklist.txt S. Ravuri]<br />
−
[/results/cover/TA.maz.ranklist.txt]
+
'''SZA''' : [https://music-ir.org/mirex/results/2009/cover/SZA.maz.ranklist.txt J. Serra et al.]<br />
 +
'''TA''' : [https://music-ir.org/mirex/results/2009/cover/TA.maz.ranklist.txt T. Ahonen]<br />
  
  
 
[[Category: Results]]
 
[[Category: Results]]

Latest revision as of 21:44, 13 May 2010

Introduction

This year, we ran Audio Cover Song (ACS) Identfication with two datasets: Mixed Collection and Sapp's Mazurka Collection.

Mixed Collection Information

This is the "original" ACS collection. Within the 1000 pieces in the Audio Cover Song database, there are embedded 30 different "cover songs" each represented by 11 different "versions" for a total of 330 audio files (16bit, monophonic, 22.05khz, wav). The "cover songs" represent a variety of genres (e.g., classical, jazz, gospel, rock, folk-rock, etc.) and the variations span a variety of styles and orchestrations.

Using each of these cover song files in turn as as the "seed/query" file, we will examine the returned lists of items for the presence of the other 10 versions of the "seed/query" file.

Sapp's Mazurka Collection Information

In addition to our original ACS dataset, we used the Mazurka.org dataset put together by Craig Sapp. We randomly chose 11 versions from 49 mazurkas and ran it as a separate ACS subtask. Systems returned a distance matrix of 539x539 from which we located the ranks of each of the associated cover versions.

General Legend

Team ID

RE = Suman Ravuri, Daniel P. W. Ellis
SZA = Joan Serra, Massimiliano Zanin, Ralph G. Andrzejak
TA = Teppo E. Ahonen

Results

Mixed Collection

Summary Results

RE SZA TA
Total number of covers identified in top 10 2046.00 2426.00 646.00
Mean number of covers identified in top 10 (average performance) 6.20 7.35 1.96
Mean (arithmetic) of Avg. Precisions 0.66 0.75 0.20
Mean rank of first correctly identified cover 2.28 6.15 29.90

download these results as csv

Number of Correct Covers at Rank X Returned in Top Ten

Rank RE SZA TA
1 285 284 118
2 271 285 108
3 245 282 85
4 246 279 80
5 228 265 58
6 208 252 61
7 196 230 46
8 160 227 33
9 139 186 32
10 68 136 25
Total 2046 2426 646

download these results as csv

Average Performance per Query Group

Group RE SZA TA
1 0.56 0.97 0.02
2 0.90 0.75 0.04
3 0.29 0.05 0.06
4 0.57 1 0.04
5 0.03 0.01 0.02
6 0.77 0.76 0.11
7 0.89 1 0.15
8 0.91 1 0.61
9 0.94 0.96 0.32
10 0.75 0.99 0.14
11 0.22 0.72 0.08
12 0.65 0.51 0.29
13 1 1 0.90
14 0.45 0.93 0.02
15 0.92 0.90 0.07
16 0.42 0.21 0.01
17 0.88 0.90 0.49
18 0.84 0.90 0.05
19 0.95 1 0.79
20 0.46 0.06 0.01
21 0.57 0.21 0.03
22 0.85 1 0.04
23 0.94 0.99 0.07
24 0.86 0.98 0.58
25 0.91 0.91 0.49
26 0.43 0.91 0.05
27 0.23 0.81 0.04
28 0.82 0.80 0.26
29 0.79 0.89 0.02
30 0.84 0.96 0.02

download these results as csv


Friedman's Test for Significant Differences

The Friedman test was run in MATLAB against the Average Precision summary data over the 30 song groups.
Command: [c,m,h,gnames] = multcompare(stats, 'ctype', 'tukey-kramer','estimate', 'friedman', 'alpha', 0.05);

TeamID TeamID Lowerbound Mean Upperbound Significance
SZA RE -0.44 0.17 0.77 FALSE
SZA TA 0.88 1.48 2.09 TRUE
RE TA 0.71 1.32 1.92 TRUE

download these results as csv


2009 coversong mixed ave performance per query group friedman mean ranks.png

Run Times

TBA

Sapp's Mazuraka Collection

Summary results

RE SZA TA
Total number of covers identified in top 10 4757.00 5165.00 2843.00
Mean number of covers identified in top 10 (average performance) 8.83 9.58 5.27
Mean (arithmetic) of Avg. Precisions 0.91 0.96 0.56
Mean rank of first correctly identified cover 1.68 1.61 5.49

download these results as csv

Number of Correct Covers at Rank X Returned in Top Ten

Rank RE SZA TA
1 522 527 451
2 522 522 413
3 522 525 368
4 523 516 320
5 518 522 316
6 513 522 271
7 495 518 235
8 452 506 195
9 415 511 160
10 285 496 114
Total 4767 5165 2843

download these results as csv

Average Performance per Query Group

Group RE SZA TA
1 1.00 1 0.61
2 1 1 0.94
3 1 1 0.72
4 0.98 0.98 0.92
5 1 1 0.02
6 0.89 1 0.47
7 1 1 0.97
8 1 1 0.45
9 1 1 0.96
10 0.92 1 0.68
11 0.97 1 0.90
12 0.85 1 0.32
13 1 1 0.31
14 0.99 1 0.26
15 0.97 1 0.80
16 0.99 1 0.34
17 0.99 1 0.05
18 0.98 1 0.81
19 1 0.94 0.86
20 1 1 0.98
21 1 1 0.49
22 0.90 1 0.97
23 1 1 0.90
24 0.75 1 0.56
25 0.99 1 0.41
26 0.71 0.73 0.08
27 0.90 0.84 0.60
28 0.58 0.45 0.45
29 0.45 0.49 0.32
30 0.93 1 0.30
31 0.91 1 0.30
32 0.94 1 0.87
33 1 1 0.74
34 0.94 1 0.28
35 0.92 1 0.40
36 0.90 1 0.75
37 0.91 1 0.27
38 0.75 1 0.05
39 0.99 1 0.87
40 1 1 0.04
41 0.89 1 0.94
42 0.91 0.99 0.47
43 0.83 1 0.30
44 0.90 1 0.18
45 0.75 1 0.03
46 1 1 0.41
47 1 1 0.91
48 0.98 1 0.72
49 0.89 0.90 0.46

download these results as csv


Friedman's Test for Significant Differences

The Friedman test was run in MATLAB against the Average Precision summary data over the 30 song groups.
Command: [c,m,h,gnames] = multcompare(stats, 'ctype', 'tukey-kramer','estimate', 'friedman', 'alpha', 0.05);

TeamID TeamID Lowerbound Mean Upperbound Significance
SZA RE 0.1764 0.6327 1.0889 TRUE
SZA TA 1.2988 1.7551 2.2114 TRUE
RE TA 0.6662 1.1224 1.5787 TRUE

download these results as csv


2009 coversong mazurkas ave performance per query group friedman mean ranks.png

Run Times

TBA

Individual Results Files

Mixed Collection

Average Precision by Query

RE : S. Ravuri, D.P.W Ellis
SZA : J. Serra et al.
TA : T. Ahonen

Average Precision by Cover Group

RE : S. Ravuri, D.P.W Ellis
SZA : J. Serra et al.
TA : T. Ahonen

Rank Lists

RE : S. Ravuri, D.P.W Ellis
SZA : J. Serra et al.
TA : T. Ahonen

Sapp's Mazurka Collection

Average Precision by Query

RE : S. Ravuri
SZA : J. Serra et al.
TA : T. Ahonen

Average Precision by Cover Group

RE : S. Ravuri
SZA : J. Serra et al.
TA : T. Ahonen

Rank Lists

RE : S. Ravuri
SZA : J. Serra et al.
TA : T. Ahonen

Retrieved from "https://music-ir.org/mirex/w/index.php?title=2009:Audio_Cover_Song_Identification_Results&oldid=6697"