2010:Audio Cover Song Identification Results

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

Sub code	Submission name	Abstract	Contributors
MHRAF1	Simbals_Cover_Songs	PDF	Benjamin Martin, Pierre Hanna, Matthias Robine, Julien Allali, Pascal Ferraro
MOD1	Applying Text-Based IR Techniques to Cover Song Identification	PDF	Nicola Montecchio, Nicola Orio, Emanuele Di Buccio
RMHAR1	Cover	PDF	Thomas Rocher, Benjamin Martin, Pierre Hanna, Julien Allali, Matthias Robine, Pascal Ferraro

Results

Mixed Collection

Summary Results

	MHRAF1	MOD1	RMHAR1
Total number of covers identified in top 10	780.00	471.00	908.00
Mean number of covers identified in top 10 (average performance)	2.36	1.43	2.75
Mean (arithmetic) of Avg. Precisions	0.24	0.15	0.29
Mean rank of first correctly identified cover	28.16	39.34	38.35

download these results as csv

Number of Correct Covers at Rank X Returned in Top Ten

Rank	MHRAF1	MOD1	RMHAR1
1	94	106	159
2	101	82	139
3	105	57	115
4	88	53	98
5	92	37	93
6	70	34	79
7	64	31	75
8	58	29	56
9	59	26	51
10	49	16	43
Total	780	471	908

download these results as csv

Average Performance per Query Group

Group	MHRAF1	MOD1	RMHAR1
1	0.25	0.01	0.04
2	0.09	0.01	0.12
3	0.01	0.08	0.02
4	0.03	0.03	0.02
5	0.02	0.01	0.01
6	0.20	0.32	0.25
7	0.87	0.02	0.89
8	0.82	0.74	0.47
9	0.47	0.21	0.62
10	0.12	0.17	0.34
11	0.36	0.02	0.40
12	0.23	0.03	0.28
13	0.68	0.18	0.98
14	0.08	0.02	0.31
15	0.38	0.02	0.98
16	0.01	0.07	0.02
17	0.75	0.15	0.55
18	0.70	0.28	0.73
19	0.28	0.31	0.80
20	0.03	0.02	0.04
21	0.11	0.02	0.30
22	0.06	0.03	0.03
23	0.42	0.71	0.15
24	0.25	0.73	0.53
25	0.56	0.02	0.63
26	0.03	0.05	0.02
27	0.01	0.38	0.01
28	0.33	0.23	0.05
29	0.11	0.01	0.14
30	0.12	0.03	0.11

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
RMHAR1	MHRAF1	-0.37	0.23	0.84	FALSE
RMHAR1	MOD1	-0.04	0.57	1.17	FALSE
MHRAF1	MOD1	-0.27	0.33	0.94	FALSE

download these results as csv

Run Times

TBA

Sapp's Mazuraka Collection

Summary results

	MHRAF1	MOD1	RMHAR1
Total number of covers identified in top 10	4071.00	3255.00	4279.00
Mean number of covers identified in top 10 (average performance)	7.55	6.04	7.94
Mean (arithmetic) of Avg. Precisions	0.79	0.63	0.82
Mean rank of first correctly identified cover	2.09	3.03	3.42

download these results as csv

Number of Correct Covers at Rank X Returned in Top Ten

Rank	MHRAF1	MOD1	RMHAR1
1	446	489	496
2	450	462	495
3	446	432	473
4	445	406	461
5	432	366	458
6	440	325	452
7	411	273	423
8	365	229	387
9	344	158	345
10	292	115	289
Total	4071	3255	4279

download these results as csv

Average Performance per Query Group

Group	MHRAF1	MOD1	RMHAR1
1	0.88	0.85	1
2	0.85	0.98	1
3	0.88	0.61	0.88
4	0.38	0.07	0.02
5	0.88	0.78	0.33
6	0.89	0.52	1
7	0.79	0.83	1
8	0.72	0.65	0.45
9	0.81	0.13	0.22
10	0.99	0.56	1
11	0.72	1.00	1
12	0.60	0.88	0.66
13	0.98	0.74	1
14	0.77	0.90	0.93
15	1	0.72	0.99
16	0.98	0.89	0.87
17	0.70	0.85	1
18	0.88	0.50	0.72
19	0.88	0.70	0.29
20	1	0.91	1
21	0.79	0.58	1
22	0.97	0.96	0.91
23	0.90	0.75	0.85
24	0.90	0.76	0.80
25	0.93	0.53	1
26	0.66	0.72	0.71
27	0.67	0.61	0.90
28	0.61	0.30	0.39
29	0.31	0.29	0.37
30	0.99	0.15	0.97
31	0.93	0.62	1
32	0.91	1	1
33	0.96	0.64	1
34	0.81	0.56	0.45
35	0.97	0.72	1
36	0.93	0.70	1
37	0.86	0.54	1
38	0.73	0.58	0.93
39	0.94	0.20	0.86
40	1	0.71	1
41	0.96	0.92	0.92
42	0.25	0.35	0.14
43	0.79	0.56	0.76
44	0.93	0.85	0.76
45	0.16	0.25	0.07
46	0.59	0.39	0.83
47	0.74	0.97	0.99
48	0.80	0.66	0.95
49	0.90	0.43	0.78

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
MHRAF1	RMHAR1	-0.5209	-0.0510	0.4188	FALSE
MHRAF1	MOD1	0.2546	0.7245	1.1944	TRUE
RMHAR1	MOD1	0.3056	0.7755	1.2454	TRUE

download these results as csv

Run Times

TBA

Individual Results Files

Mixed Collection

Average Precision by Query

MHRAF1 : Benjamin Martin et al.
MOD1 : Nicola Montecchio et al.
RMHAR1 : Thomas Rocher et al.

Rank Lists

MHRAF1 : Benjamin Martin et al.
MOD1 : Nicola Montecchio et al.
RMHAR1 : Thomas Rocher et al.

Sapp's Mazurka Collection

Average Precision by Query

MHRAF1 : Benjamin Martin et al.
MOD1 : Nicola Montecchio et al.
RMHAR1 : Thomas Rocher et al.

Rank Lists

MHRAF1 : Benjamin Martin et al.
MOD1 : Nicola Montecchio et al.
RMHAR1 : Thomas Rocher et al.