Difference between revisions of "2017:Audio Cover Song Identification Results"

From MIREX Wiki
(Results)
Line 44: Line 44:
  
 
====Summary Results====
 
====Summary Results====
<csv p=2>2016/acs/coversong1000.summary.csv</csv>
+
<csv p=2>2017/acs/coversong1000.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>2016/acs/coversong1000.toptendist.csv</csv>
+
<csv>2017/acs/coversong1000.toptendist.csv</csv>
  
 
====Average Performance per Query Group====
 
====Average Performance per Query Group====
<csv p=2>2016/acs/coversong1000.precision.groups.csv</csv>
+
<csv p=2>2017/acs/coversong1000.precision.groups.csv</csv>
  
 
===Sapp's Mazuraka Collection===
 
===Sapp's Mazuraka Collection===
  
 
====Summary results====
 
====Summary results====
<csv p=2>2016/acs/mazurkas.summary.csv</csv>
+
<csv p=2>2017/acs/mazurkas.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>2016/acs/mazurkas.toptendist.csv</csv>
+
<csv>2017/acs/mazurkas.toptendist.csv</csv>
  
 
====Average Performance per Query Group====
 
====Average Performance per Query Group====
<csv p=2>2016/acs/mazurkas.precision.groups.csv</csv>
+
<csv p=2>2017/acs/mazurkas.precision.groups.csv</csv>
  
 
==Individual Results Files==
 
==Individual Results Files==

Revision as of 01:15, 23 October 2017

Introduction

Audio Cover Song (ACS) Identification was evaluated against 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 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

Sub code Submission name Abstract Contributors
CL1 CL1 PDF Ning Chen, Hai-Dong Xiao
YWJ2 YWJ2 PDF Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang
YWJ3 YWJ3 PDF Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang
CYWW1 CYWW1 PDF Chuan-Yau Chan, Ming-Chi Yen, Ju-Chiang Wang, Hsin-Min Wang
SYBK1 SYBK1 PDF Diego Silva, Chin-Chia Michael Yeh, Gustavo Enrique de Almeida Prado Alves Batista, Eamonn Keogh


Results

Mixed Collection

Summary Results

CL1 CT2 HKL1Fast1 HKL1Fast21
Total number of covers identified in top 10 2104.00 1967.00 1496.00 1488.00
Mean number of covers identified in top 10 (average performance) 6.38 5.96 4.53 4.51
Mean (arithmetic) of Avg. Precisions 0.65 0.62 0.47 0.47
Mean rank of first correctly identified cover 52.40 18.80 15.10 14.70

download these results as csv

Number of Correct Covers at Rank X Returned in Top Ten

Rank CL1 CT2 HKL1Fast1 HKL1Fast21
1 251 289 213 214
2 241 264 186 185
3 240 256 186 185
4 230 250 156 157
5 233 223 152 154
6 215 215 158 154
7 207 166 141 137
8 202 156 130 134
9 186 103 103 108
10 99 45 71 60
Total 2104 1967 1496 1488

download these results as csv

Average Performance per Query Group

Group CL1 CT2 HKL1Fast1 HKL1Fast21
1 0.82 0.76 0.82 0.82
2 0.04 0.20 0.05 0.05
3 0.71 0.40 0.55 0.55
4 0.82 0.83 0.68 0.67
5 0.84 0.83 0.23 0.23
6 0.99 0.68 0.98 0.98
7 0.46 0.61 0.24 0.24
8 0.44 0.71 0.24 0.23
9 0.14 0.53 0.03 0.03
10 0.42 0.31 0.60 0.60
11 0.18 0.35 0.16 0.16
12 0.80 0.82 0.74 0.74
13 0.82 0.51 0.98 0.98
14 0.47 0.70 0.12 0.12
15 1 0.82 0.90 0.90
16 0.82 0.82 0.43 0.43
17 0.33 0.39 0.23 0.23
18 0.98 0.86 0.46 0.46
19 0.74 0.58 0.64 0.64
20 0.66 0.29 0.28 0.28
21 0.57 0.58 0.06 0.06
22 0.81 0.68 0.18 0.17
23 0.64 0.66 0.31 0.32
24 1.00 0.96 0.96 0.95
25 0.16 0.83 0.09 0.09
26 0.32 0.51 0.33 0.32
27 0.82 0.53 0.67 0.67
28 0.72 0.49 0.31 0.30
29 1 0.75 0.96 0.97
30 0.91 0.72 0.87 0.87

download these results as csv

Sapp's Mazuraka Collection

Summary results

CL1 CT2 HKL1Fast1 HKL1Fast21
Total number of covers identified in top 10 5139.00 4950.00 4773.00 4778.00
Mean number of covers identified in top 10 (average performance) 9.53 9.18 8.86 8.86
Mean (arithmetic) of Avg. Precisions 0.95 0.92 0.90 0.90
Mean rank of first correctly identified cover 1.54 3.15 1.96 1.96

download these results as csv

Number of Correct Covers at Rank X Returned in Top Ten

Rank CL1 CT2 HKL1Fast1 HKL1Fast21
1 524 517 519 518
2 526 514 516 522
3 513 508 520 515
4 516 515 506 504
5 519 511 495 498
6 520 504 484 486
7 516 505 478 476
8 504 485 445 446
9 517 479 415 419
10 484 412 395 394
Total 5139 4950 4773 4778

download these results as csv

Average Performance per Query Group

Group CL1 CT2 HKL1Fast1 HKL1Fast21
1 1 1 0.97 0.97
2 1 1 1 1
3 1 0.85 0.98 0.98
4 1 0.67 0.83 0.83
5 1 0.83 0.73 0.73
6 1 1 0.78 0.79
7 1 0.99 1 1
8 1 0.84 0.60 0.60
9 1 0.76 0.76 0.75
10 1 1 0.97 0.97
11 1 1 1 1
12 1 1 1 1
13 1 1 1 1
14 1 1 0.97 0.96
15 1 1 0.97 0.97
16 1 1 0.97 0.97
17 1 1 1 1
18 1 1 1 1
19 1 0.99 1 1
20 1 1 1 1
21 1 1 0.98 0.98
22 1 1 0.98 0.98
23 1 1 0.97 0.97
24 1 1 1 1
25 1 1 1 1
26 0.65 0.71 0.64 0.64
27 0.77 0.76 0.77 0.77
28 0.48 0.42 0.38 0.38
29 0.39 0.34 0.40 0.39
30 1 1 0.98 0.98
31 0.82 1 0.80 0.81
32 1 1 0.94 0.94
33 1 1 1 1
34 1 0.92 0.90 0.91
35 1 1 0.98 0.98
36 1 1 1 1
37 1 0.91 0.99 0.99
38 0.94 1 1 1
39 1 1 1.00 1.00
40 1 1 0.99 0.98
41 1 1 1 1
42 0.93 0.54 0.66 0.66
43 1 0.99 0.73 0.74
44 1 0.94 1 1
45 0.82 1.00 0.93 0.93
46 1 0.99 0.72 0.71
47 1 1 1 1
48 1 1 0.75 0.75
49 0.83 0.82 0.82 0.82

download these results as csv

Individual Results Files

Mixed Collection

Average Precision by Query

CL1 : Ning Chen, Hai-Dong Xiao

CYWW1 : Chuan-Yau Chan, Ming-Chi Yen, Ju-Chiang Wang, Hsin-Min Wang

SYBK1 : Diego Silva, Chin-Chia Michael Yeh, Gustavo Enrique de Almeida Prado Alves Batista, Eamonn Keogh

YWJ2 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang

YWJ3 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang


Rank Lists

CL1 : Ning Chen, Hai-Dong Xiao

CYWW1 : Chuan-Yau Chan, Ming-Chi Yen, Ju-Chiang Wang, Hsin-Min Wang

SYBK1 : Diego Silva, Chin-Chia Michael Yeh, Gustavo Enrique de Almeida Prado Alves Batista, Eamonn Keogh

YWJ2 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang

YWJ3 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang

Mazurkas

Average Precision by Query

CL1 : Ning Chen, Hai-Dong Xiao

CYWW1 : Chuan-Yau Chan, Ming-Chi Yen, Ju-Chiang Wang, Hsin-Min Wang

SYBK1 : Diego Silva, Chin-Chia Michael Yeh, Gustavo Enrique de Almeida Prado Alves Batista, Eamonn Keogh

YWJ2 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang

YWJ3 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang


Rank Lists

CL1 : Ning Chen, Hai-Dong Xiao

CYWW1 : Chuan-Yau Chan, Ming-Chi Yen, Ju-Chiang Wang, Hsin-Min Wang

SYBK1 : Diego Silva, Chin-Chia Michael Yeh, Gustavo Enrique de Almeida Prado Alves Batista, Eamonn Keogh

YWJ2 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang

YWJ3 : Ming-Chi Yen, Hsin-Min Wang, Jyh-Shing Roger Jang