2024:Polyphonic Transcription - Revision history

Yujiayan: /* Submission */

2024-10-04T21:28:45Z

‎Submission

2024-10-04T18:49:53Z

2024-09-16T06:07:23Z

‎Introduction

2024-09-15T18:55:22Z

‎Event-level metrics

2024-09-15T18:54:52Z

‎Event-level metrics

2024-09-15T18:13:23Z

‎Introduction

2024-09-15T18:13:05Z

‎Introduction

2024-09-15T18:12:42Z

‎Introduction

2024-09-15T18:12:10Z

‎Introduction

2024-09-15T18:11:44Z

‎Introduction

@@ Line 120: / Line 120: @@
 └── teamA.pdf (extended abstract, 2-4 pages, or the published paper about the system)
 </pre>
 == Questions or Suggestions ==

@@ Line 97: / Line 97: @@
 #* Example: "maestro_no_ext", "MAPS_no_ext", "SMD_no_ext"
 # Similarly, if you system does not output pedals, add postfix "_no_pedal" to the folder names
-# Include an extended abstract for briefly describe the design of your system. Format: PDF of 2-4 pages.
+# Include an extended abstract for briefly describe the design of your system. Format: PDF of 2-4 pages. '''If it comes from a published paper, you can choose to include that paper directly'''
 # Put all folders into a single .zip file, name it with the identifier you want for your system.
@@ Line 118: / Line 118: @@
 │   ├── Bach_BWV849-02_001_20090916-SMD.mid
 │   └── ...
-└── teamA.pdf (extended abstract, 2-4 pages)
+└── teamA.pdf (extended abstract, 2-4 pages, or the published paper about the system)
 </pre>

@@ Line 1: / Line 1: @@
 == Introduction ==
-We are introducing polyphonic transcription as a MIREX task for the first time. This year's focus is on '''piano transcription''' (audio-to-MIDI). For piano transcription, our task is designed to be compatible with most published papers on the topic (e.g., [[#References| papers in the reference list]]), aiming to standardize evaluation in the field, fixing the pervasive error in evaluation datasets beyond MAESTRO. (i.e., [[#Additional procedure for mitigating systematic biases in specific evaluation datasets| fixing the alignment bias]]).
+We are introducing polyphonic transcription as a MIREX task for the first time. This year's focus is on '''piano transcription''' (audio-to-MIDI). For piano transcription, our task is designed to be compatible with most published papers on the topic (e.g., [[#References| papers in the reference list]]), aiming to standardize evaluation in the field, fixing the pervasive error in published papers of using evaluation datasets beyond MAESTRO. (i.e., [[#Additional procedure for mitigating systematic biases in specific evaluation datasets| fixing the alignment bias]]).
 == Task Description ==

@@ Line 69: / Line 69: @@
 # '''Note''': A note prediction is considered correct if its estimated onset is within 50ms of the corresponding ground truth onset
 # '''Note with offset''': In addition to the onset criterion, the offset of the estimated note must be within 50ms or 20% of the note duration from the ground truth offset, whichever is greater
-# '''Note with offset and velocity''': In addition to onset and offset criteria, the normalized estimated midi velocity must be within 0.1 of the ground truth velocity. This normalized estimated velocity is described in (Hawthorne et al 2018) that uses linear regression to rescale and shift the estimated velocity to match the groudtruth.
+# '''Note with offset and velocity''': In addition to onset and offset criteria, the normalized estimated midi velocity must be within 10% of the ground truth velocity. This normalized estimated velocity is described in (Hawthorne et al 2018) that uses linear regression to rescale and shift the estimated velocity to match the groudtruth.
 In addition to these classification based metrics, we also compute the mean onset/offset deviations (in milliseconds) for matched notes between predicted and ground truth notes.

@@ Line 69: / Line 69: @@
 # '''Note''': A note prediction is considered correct if its estimated onset is within 50ms of the corresponding ground truth onset
 # '''Note with offset''': In addition to the onset criterion, the offset of the estimated note must be within 50ms or 20% of the note duration from the ground truth offset, whichever is greater
-# '''Note with offset and velocity''': In addition to onset and offset criteria, the normalized estimated midi velocity must be within 0.1 of the normalized ground truth velocity. This normalized estimated velocity is described in (Hawthorne et al 2018) that uses linear regression to rescale and shift the estimated velocity to match the groudtruth.
+# '''Note with offset and velocity''': In addition to onset and offset criteria, the normalized estimated midi velocity must be within 0.1 of the ground truth velocity. This normalized estimated velocity is described in (Hawthorne et al 2018) that uses linear regression to rescale and shift the estimated velocity to match the groudtruth.
 In addition to these classification based metrics, we also compute the mean onset/offset deviations (in milliseconds) for matched notes between predicted and ground truth notes.