2020:Singing Transcription from Polyphonic Music - Revision history

Jywang: /* Description */

2020-08-25T13:15:08Z

‎Description

Jywang: /* Question? */

2020-08-25T13:12:25Z

‎Question?

Jywang: /* Evaluation */

2020-08-25T10:45:23Z

‎Evaluation

Jywang: /* Evaluation */

2020-08-25T10:45:00Z

‎Evaluation

Jywang at 09:55, 25 August 2020

2020-08-25T09:55:37Z

Jywang: /* Time limits */

2020-08-25T06:21:04Z

‎Time limits

Jywang at 06:20, 25 August 2020

2020-08-25T06:20:05Z

Jywang: /* Time limits */

2020-08-25T06:05:37Z

‎Time limits

Jywang: /* Time limits */

2020-08-25T06:05:08Z

‎Time limits

Jywang: /* Description */

2020-08-25T06:03:42Z

‎Description

@@ Line 8: / Line 8: @@
 Besides, it's worth noting that the definition of “singing transcription” is not quite specific. Some researches [1][2] regard this term as the task of “transcribing '''polyphonic music''' into notes”, but other researches [3][4] seems to regard this term as the task of “transcribing '''monophonic signals without accompaniment''' into notes”, since both of them [3][4] created monophonic datasets (without accompaniments) for “automatic singing transcription”.
-Therefore, to make the name of the task clear, we name the task that “transcribing '''polyphonic music''' that contains only '''monophonic vocal''' into notes” as '''singing transcription from polyphonic music'''.
+Therefore, to make the name more specific, we call the task that “transcribing '''polyphonic music''' that contains only '''monophonic vocal''' into notes” as '''singing transcription from polyphonic music'''.
 == Data ==

@@ Line 70: / Line 70: @@
 If you have any question about this task or the datasets, please feel free to send us an email: b06902046@ntu.edu.tw (Jun-You Wang) or roger.jang@gmail.com (Jyh-Shing Roger Jang).
-Since this is a new MIREX task, we desperately need your help to make everything get on track.
+Since this is a new MIREX task, we are eagerly waiting for your help to make everything get on track.
 == Submission deadline ==

@@ Line 34: / Line 34: @@
 In fact, the COnPOff metric is the same as the evaluation metrics of “note tracking” subtask in MIREX [https://www.music-ir.org/mirex/wiki/2020:Multiple_Fundamental_Frequency_Estimation_%26_Tracking “Multiple Fundamental Frequency Estimation & Tracking” task]. The only difference is that the onset threshold is set to 100ms instead of 50ms due to the difficulty of labeling (and maybe, transcribing) onset of singing voice.
-A simple evaluation script can be downloaded from [https://drive.google.com/file/d/1Uw-MQA14XGypwXaADdvV9Hg_URsWD4WV here].
+A simple evaluation code can be downloaded from [https://drive.google.com/file/d/1Uw-MQA14XGypwXaADdvV9Hg_URsWD4WV here].
 == Submission Format ==

← Older revision		Revision as of 10:45, 25 August 2020
Line 33:		Line 33:

	In fact, the COnPOff metric is the same as the evaluation metrics of “note tracking” subtask in MIREX [https://www.music-ir.org/mirex/wiki/2020:Multiple_Fundamental_Frequency_Estimation_%26_Tracking “Multiple Fundamental Frequency Estimation & Tracking” task]. The only difference is that the onset threshold is set to 100ms instead of 50ms due to the difficulty of labeling (and maybe, transcribing) onset of singing voice.		In fact, the COnPOff metric is the same as the evaluation metrics of “note tracking” subtask in MIREX [https://www.music-ir.org/mirex/wiki/2020:Multiple_Fundamental_Frequency_Estimation_%26_Tracking “Multiple Fundamental Frequency Estimation & Tracking” task]. The only difference is that the onset threshold is set to 100ms instead of 50ms due to the difficulty of labeling (and maybe, transcribing) onset of singing voice.
		+
		+	A simple evaluation script can be downloaded from [https://drive.google.com/file/d/1Uw-MQA14XGypwXaADdvV9Hg_URsWD4WV here].

	== Submission Format ==		== Submission Format ==

@@ Line 14: / Line 14: @@
 === RWC Music Database : Popular Music (RWC-MDB-P) ===
 We can use the “Popular Music Database” part of RWC database [5] for this task. RWC-MDB-P consists of 100 songs and annotations in MIDI format (AIST annotation). The database is available at https://staff.aist.go.jp/m.goto/RWC-MDB/rwc-mdb-p.html, and the annotations are available at https://staff.aist.go.jp/m.goto/RWC-MDB/AIST-Annotation. By excluding 6 songs (No. 3, 5, 8, 10, 23, and 66) with multiple singers (the melody part is not monophonic), the remaining 94 songs can be used for this task.
 === Cmedia dataset ===
-This dataset consists of 200 Youtube links of pop songs (most of them are Chinese songs), together with their groundtruth files of vocal transcription. We will release 100 of the dataset as the open set for training/validation, and use the other 100 as the hidden set for test.
+This dataset consists of 200 Youtube links of pop songs (most of them are Chinese songs), together with their groundtruth files of vocal transcription. We will release 100 of them as the open set for training/validation (training set), and use the other 100 as the hidden set for test.
 The '''training''' set can be downloaded [https://drive.google.com/file/d/15b298vSP9cPP8qARQwa2X_0dbzl6_Eu7/ here]. We strongly suggest participants to use this dataset as training set (if your algorithm is data-driven), since the property of Cmedia training set is close to Cmedia hidden set.

@@ Line 65: / Line 65: @@
 If the algorithm cannot transcribe all 200 songs on time, it's still OK. However, the algorithm should at least transcribe 100 songs (from Cmedia hidden set) within time limit, otherwise no evaluation result can be reported.
-The algorithm will be executed on a computer with 64GB RAM and one NVIDIA GEFORCE GTX 1080Ti GPU.
+The algorithm will be executed on a computer with 64GB memory and one NVIDIA GEFORCE GTX 1080Ti GPU.
 == Question? ==

@@ Line 61: / Line 61: @@
 === Time limits ===
-The time limit is 24 hours. In 24 hours, the algorithm should transcribe all 200 songs in RWC dataset and Cmedia dataset, of which the total duration is about 14hr.
+The time limit is 24 hours. In 24 hours, the algorithm should transcribe all 200 songs in Cmedia dataset, of which the total duration is about 14hr.
 If the algorithm cannot transcribe all 200 songs on time, it's still OK. However, the algorithm should at least transcribe 100 songs (from Cmedia hidden set) within time limit, otherwise no evaluation result can be reported.

← Older revision		Revision as of 06:05, 25 August 2020
Line 64:		Line 64:

	If the algorithm cannot transcribe all 200 songs on time, it's still OK. However, the algorithm should at least transcribe 100 songs (from Cmedia hidden set) within time limit, otherwise no evaluation result can be reported.		If the algorithm cannot transcribe all 200 songs on time, it's still OK. However, the algorithm should at least transcribe 100 songs (from Cmedia hidden set) within time limit, otherwise no evaluation result can be reported.
		+
	The algorithm will be executed on a computer with 64GB RAM and 1 NVIDIA GEFORCE GTX 1080Ti GPU.		The algorithm will be executed on a computer with 64GB RAM and 1 NVIDIA GEFORCE GTX 1080Ti GPU.

@@ Line 6: / Line 6: @@
 This task is different from “audio melody extraction”, since the target of melody extraction is to determine pitch for each frame, while singing transcription is aimed to determine notes of vocal part in music.
-Besides, it's worth noting that the definition of “singing transcription” is not quite specific. Some researches [1][2] regard this term as the task of “transcribing '''polyphonic music''' into notes”, but other researches [3][4] seems to regard this term as the task of “transcribing '''monophonic signals without accompaniment''' into notes”, since both of them created monophonic datasets (without accompaniments) for “automatic singing transcription”.
+Besides, it's worth noting that the definition of “singing transcription” is not quite specific. Some researches [1][2] regard this term as the task of “transcribing '''polyphonic music''' into notes”, but other researches [3][4] seems to regard this term as the task of “transcribing '''monophonic signals without accompaniment''' into notes”, since both of them [3][4] created monophonic datasets (without accompaniments) for “automatic singing transcription”.
 Therefore, to make the name of the task clear, we name the task that “transcribing '''polyphonic music''' that contains only '''monophonic vocal''' into notes” as '''singing transcription from polyphonic music'''.