MIREX Wiki - User contributions [en]

2026:Music Performance Difficulty Prediction

2026-07-01T09:36:49Z

Huanz: /* Training Datasets */

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image-based, and audio-based modalities (Ramoneda et al., 2022–2025). Among these directions, audio-based difficulty estimation remains the most recent and least explored. Unlike symbolic or score-based approaches, audio-only methods do not require machine-readable scores, substantially widening the applicability of difficulty-prediction systems for music libraries, educational platforms, teachers, students, and recommendation systems.

Despite this growing interest, the field still lacks a standardized community benchmark for audio-based difficulty prediction. Existing studies rely on heterogeneous datasets, incompatible grading systems, and non-comparable evaluation protocols. This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty score.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': a real-valued difficulty score per recording. The score is treated as an ordering — its absolute scale, range, or number of categories does not need to match the (hidden) evaluation scale.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized score rendering from a fixed soundfont — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

There is no inference-time limit per piece, but the full test set must be processed within a 24-hour wall-clock budget on a single GPU.

== Significance of the Task ==

This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio, featuring:

* A consistent ordinal evaluation framework
* A fully private held-out test set curated specifically for this task
* Standardized ordinal regression and ranking metrics
* Composer-disjoint evaluation splits designed to measure true generalization beyond repertoire memorization
* A hidden pedagogical difficulty scale validated by expert pedagogues and musicologists

Beyond benchmarking, the task aims to lower the entry barrier for researchers working on audio representation learning, multimodal learning, and music understanding, while fostering a growing research community at the intersection of MIR, pedagogy, and computational music education.

== Evaluation Criteria ==

During development, participants are encouraged to evaluate their systems on public datasets such as PSyllabus using standard ordinal-regression metrics, including:

* Mean Squared Error (MSE)
* Accuracy within one level (Acc±1)
* Balanced accuracy
* Spearman's ρ

These metrics are useful for model selection and comparison on publicly annotated datasets with explicit grade structures.

However, because the official evaluation set uses a '''hidden ordinal pedagogical scale''' whose granularity and number of levels are not disclosed, the shared-task ranking is based exclusively on '''Kendall's Tau-c''', which measures agreement between predicted and reference orderings without assuming fixed distances between categories.

In addition to the official ranking metric, organizers may report supplementary analyses such as per-composer performance to better understand cross-repertoire generalization and robustness across stylistic domains.

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/14794592 Psyllabus]''' (Ramoneda et al., 2025): 7,901 audio recordings of classical piano pieces with 11-level pedagogical grades, together with mappings to multiple international grading systems. We specifically encourage participants to leverage the multiple ranking annotations available in PSyllabus (13 grading/ranking systems in total), not only the default labels, in order to study and improve generalization across heterogeneous pedagogical traditions and difficulty scales.
* '''[https://zenodo.org/records/8037327 Can I Play It? (CIPI)]''' (Ramoneda et al., 2024): 652 MusicXML piano pieces with 9-level Henle annotations, useful for participants interested in transcription-then-classification pipelines or distant supervision approaches.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''' (Ramoneda et al., 2022) and other publicly available piano difficulty datasets and pedagogical collections.

Participants may use any additional external data, provided that no part of the held-out evaluation repertoire is used directly or indirectly for training. To minimize contamination risks, a list of forbidden test pieces and composers will be published together with the task call.

Please describe in your technical report:
* Dataset names and sources
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To address the central concern of label trust and contamination, the held-out test set is constructed as a '''fully private benchmark''' curated specifically for this task. The dataset is assembled and annotated by expert piano pedagogues, with all annotations independently validated by a pedagogue and musicologist expert. Difficulty labels are assigned using an internal ordinal scale inspired by established pedagogical curricula and examination systems, but the exact mapping, granularity, and total number of levels are intentionally undisclosed to participants.

* The repertoire is selected from sources outside existing public piano difficulty datasets, including contemporary works, underrepresented composers, and recent pedagogical material not present in commonly used corpora. This design minimizes overlap with publicly available graded datasets and reduces the possibility of memorization or contamination effects.
* All annotations undergo independent expert review and adjudication before inclusion in the benchmark. Neither the raw labels nor the precise scale definition are released publicly.
* The held-out scores, audio recordings, and annotations remain private to the task organizers throughout and after the evaluation. Participants submit executable systems (Docker containers preferred), which are run by the organizers on the hidden evaluation set.

Because the underlying difficulty scale is ordinal and intentionally hidden, system performance is evaluated primarily through rank correlation rather than exact class prediction. The official ranking metric is '''Kendall's Tau-c''', which measures agreement between predicted and reference difficulty orderings while remaining robust to unknown category spacing and differing numbers of ordinal levels.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is a single real-valued difficulty score. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

* '''July 1, 2026''': Submission system opens
* '''Oct 1, 2026''': Submission deadline
* '''Oct 15, 2026''': Results published

== Long-term Plan ==

We are committed to maintaining the task for at least three iterations. We will collaborate with the broader piano-difficulty research community to keep label-space conventions consistent across this task and parallel evaluation efforts. The hidden held-out test set is designed to grow incrementally across editions, with prior-year items optionally re-used subject to confirmation that participants have not gained access.

Given the field's relatively small size, we propose running this task annually for the first two years to bootstrap community engagement, and biennially thereafter. A symbolic-input track may be added in future editions if community interest warrants it.

== Organizers ==

* '''Pedro Ramoneda''' (Songscription) — [mailto:pedro@songscription.ai pedro@songscription.ai]
* '''Huan Zhang''' (Clefer) — [mailto:huan@clefer.com huan@clefer.com]

== References ==

* Ramoneda, P., Jeong, D., Eremenko, V., Tamer, N. C., Miron, M., & Serra, X. (2024). Combining piano performance dimensions for score difficulty classification. ''Expert Systems with Applications'', 238, 1–16.
* Ramoneda, P., Lee, M., Jeong, D., Valero-Mas, J. J., & Serra, X. (2025). Can audio reveal music performance difficulty? Insights from the Piano Syllabus Dataset.
* Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X., & Miron, M. (2022). Score difficulty analysis for piano performance education based on fingering. In ''ICASSP 2022 – IEEE International Conference on Acoustics, Speech and Signal Processing'' (pp. 201–205). IEEE.

2026:Music Performance Difficulty Prediction

2026-07-01T09:34:58Z

Huanz: /* Training Datasets */

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image-based, and audio-based modalities (Ramoneda et al., 2022–2025). Among these directions, audio-based difficulty estimation remains the most recent and least explored. Unlike symbolic or score-based approaches, audio-only methods do not require machine-readable scores, substantially widening the applicability of difficulty-prediction systems for music libraries, educational platforms, teachers, students, and recommendation systems.

Despite this growing interest, the field still lacks a standardized community benchmark for audio-based difficulty prediction. Existing studies rely on heterogeneous datasets, incompatible grading systems, and non-comparable evaluation protocols. This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty score.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': a real-valued difficulty score per recording. The score is treated as an ordering — its absolute scale, range, or number of categories does not need to match the (hidden) evaluation scale.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized score rendering from a fixed soundfont — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

There is no inference-time limit per piece, but the full test set must be processed within a 24-hour wall-clock budget on a single GPU.

== Significance of the Task ==

This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio, featuring:

* A consistent ordinal evaluation framework
* A fully private held-out test set curated specifically for this task
* Standardized ordinal regression and ranking metrics
* Composer-disjoint evaluation splits designed to measure true generalization beyond repertoire memorization
* A hidden pedagogical difficulty scale validated by expert pedagogues and musicologists

Beyond benchmarking, the task aims to lower the entry barrier for researchers working on audio representation learning, multimodal learning, and music understanding, while fostering a growing research community at the intersection of MIR, pedagogy, and computational music education.

== Evaluation Criteria ==

During development, participants are encouraged to evaluate their systems on public datasets such as PSyllabus using standard ordinal-regression metrics, including:

* Mean Squared Error (MSE)
* Accuracy within one level (Acc±1)
* Balanced accuracy
* Spearman's ρ

These metrics are useful for model selection and comparison on publicly annotated datasets with explicit grade structures.

However, because the official evaluation set uses a '''hidden ordinal pedagogical scale''' whose granularity and number of levels are not disclosed, the shared-task ranking is based exclusively on '''Kendall's Tau-c''', which measures agreement between predicted and reference orderings without assuming fixed distances between categories.

In addition to the official ranking metric, organizers may report supplementary analyses such as per-composer performance to better understand cross-repertoire generalization and robustness across stylistic domains.

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/12783403]''' (Ramoneda et al., 2025): 7,901 audio recordings of classical piano pieces with 11-level pedagogical grades, together with mappings to multiple international grading systems. We specifically encourage participants to leverage the multiple ranking annotations available in PSyllabus (13 grading/ranking systems in total), not only the default labels, in order to study and improve generalization across heterogeneous pedagogical traditions and difficulty scales.
* '''[https://zenodo.org/records/8037327]''' (Ramoneda et al., 2024): 652 MusicXML piano pieces with 9-level Henle annotations, useful for participants interested in transcription-then-classification pipelines or distant supervision approaches.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''' (Ramoneda et al., 2022) and other publicly available piano difficulty datasets and pedagogical collections.

Participants may use any additional external data, provided that no part of the held-out evaluation repertoire is used directly or indirectly for training. To minimize contamination risks, a list of forbidden test pieces and composers will be published together with the task call.

Please describe in your technical report:
* Dataset names and sources
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To address the central concern of label trust and contamination, the held-out test set is constructed as a '''fully private benchmark''' curated specifically for this task. The dataset is assembled and annotated by expert piano pedagogues, with all annotations independently validated by a pedagogue and musicologist expert. Difficulty labels are assigned using an internal ordinal scale inspired by established pedagogical curricula and examination systems, but the exact mapping, granularity, and total number of levels are intentionally undisclosed to participants.

* The repertoire is selected from sources outside existing public piano difficulty datasets, including contemporary works, underrepresented composers, and recent pedagogical material not present in commonly used corpora. This design minimizes overlap with publicly available graded datasets and reduces the possibility of memorization or contamination effects.
* All annotations undergo independent expert review and adjudication before inclusion in the benchmark. Neither the raw labels nor the precise scale definition are released publicly.
* The held-out scores, audio recordings, and annotations remain private to the task organizers throughout and after the evaluation. Participants submit executable systems (Docker containers preferred), which are run by the organizers on the hidden evaluation set.

Because the underlying difficulty scale is ordinal and intentionally hidden, system performance is evaluated primarily through rank correlation rather than exact class prediction. The official ranking metric is '''Kendall's Tau-c''', which measures agreement between predicted and reference difficulty orderings while remaining robust to unknown category spacing and differing numbers of ordinal levels.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is a single real-valued difficulty score. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

* '''July 1, 2026''': Submission system opens
* '''Oct 1, 2026''': Submission deadline
* '''Oct 15, 2026''': Results published

== Long-term Plan ==

We are committed to maintaining the task for at least three iterations. We will collaborate with the broader piano-difficulty research community to keep label-space conventions consistent across this task and parallel evaluation efforts. The hidden held-out test set is designed to grow incrementally across editions, with prior-year items optionally re-used subject to confirmation that participants have not gained access.

Given the field's relatively small size, we propose running this task annually for the first two years to bootstrap community engagement, and biennially thereafter. A symbolic-input track may be added in future editions if community interest warrants it.

== Organizers ==

* '''Pedro Ramoneda''' (Songscription) — [mailto:pedro@songscription.ai pedro@songscription.ai]
* '''Huan Zhang''' (Clefer) — [mailto:huan@clefer.com huan@clefer.com]

== References ==

* Ramoneda, P., Jeong, D., Eremenko, V., Tamer, N. C., Miron, M., & Serra, X. (2024). Combining piano performance dimensions for score difficulty classification. ''Expert Systems with Applications'', 238, 1–16.
* Ramoneda, P., Lee, M., Jeong, D., Valero-Mas, J. J., & Serra, X. (2025). Can audio reveal music performance difficulty? Insights from the Piano Syllabus Dataset.
* Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X., & Miron, M. (2022). Score difficulty analysis for piano performance education based on fingering. In ''ICASSP 2022 – IEEE International Conference on Acoustics, Speech and Signal Processing'' (pp. 201–205). IEEE.

2026:Music Performance Difficulty Prediction

2026-05-31T14:25:12Z

Huanz: /* Timeline */

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image-based, and audio-based modalities (Ramoneda et al., 2022–2025). Among these directions, audio-based difficulty estimation remains the most recent and least explored. Unlike symbolic or score-based approaches, audio-only methods do not require machine-readable scores, substantially widening the applicability of difficulty-prediction systems for music libraries, educational platforms, teachers, students, and recommendation systems.

Despite this growing interest, the field still lacks a standardized community benchmark for audio-based difficulty prediction. Existing studies rely on heterogeneous datasets, incompatible grading systems, and non-comparable evaluation protocols. This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty score.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': a real-valued difficulty score per recording. The score is treated as an ordering — its absolute scale, range, or number of categories does not need to match the (hidden) evaluation scale.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized score rendering from a fixed soundfont — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

There is no inference-time limit per piece, but the full test set must be processed within a 24-hour wall-clock budget on a single GPU.

== Significance of the Task ==

This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio, featuring:

* A consistent ordinal evaluation framework
* A fully private held-out test set curated specifically for this task
* Standardized ordinal regression and ranking metrics
* Composer-disjoint evaluation splits designed to measure true generalization beyond repertoire memorization
* A hidden pedagogical difficulty scale validated by expert pedagogues and musicologists

Beyond benchmarking, the task aims to lower the entry barrier for researchers working on audio representation learning, multimodal learning, and music understanding, while fostering a growing research community at the intersection of MIR, pedagogy, and computational music education.

== Evaluation Criteria ==

During development, participants are encouraged to evaluate their systems on public datasets such as PSyllabus using standard ordinal-regression metrics, including:

* Mean Squared Error (MSE)
* Accuracy within one level (Acc±1)
* Balanced accuracy
* Spearman's ρ

These metrics are useful for model selection and comparison on publicly annotated datasets with explicit grade structures.

However, because the official evaluation set uses a '''hidden ordinal pedagogical scale''' whose granularity and number of levels are not disclosed, the shared-task ranking is based exclusively on '''Kendall's Tau-c''', which measures agreement between predicted and reference orderings without assuming fixed distances between categories.

In addition to the official ranking metric, organizers may report supplementary analyses such as per-composer performance to better understand cross-repertoire generalization and robustness across stylistic domains.

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/12783403 PSyllabus]''' (Ramoneda et al., 2025): 7,901 audio recordings of classical piano pieces with 11-level pedagogical grades, together with mappings to multiple international grading systems. We specifically encourage participants to leverage the multiple ranking annotations available in PSyllabus (13 grading/ranking systems in total), not only the default labels, in order to study and improve generalization across heterogeneous pedagogical traditions and difficulty scales.
* '''[https://zenodo.org/records/8037327 CIPI (Can I Play It?)]''' (Ramoneda et al., 2024): 652 MusicXML piano pieces with 9-level Henle annotations, useful for participants interested in transcription-then-classification pipelines or distant supervision approaches.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''' (Ramoneda et al., 2022) and other publicly available piano difficulty datasets and pedagogical collections.

Participants may use any additional external data, provided that no part of the held-out evaluation repertoire is used directly or indirectly for training. To minimize contamination risks, a list of forbidden test pieces and composers will be published together with the task call.

Please describe in your technical report:
* Dataset names and sources
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To address the central concern of label trust and contamination, the held-out test set is constructed as a '''fully private benchmark''' curated specifically for this task. The dataset is assembled and annotated by expert piano pedagogues, with all annotations independently validated by a pedagogue and musicologist expert. Difficulty labels are assigned using an internal ordinal scale inspired by established pedagogical curricula and examination systems, but the exact mapping, granularity, and total number of levels are intentionally undisclosed to participants.

* The repertoire is selected from sources outside existing public piano difficulty datasets, including contemporary works, underrepresented composers, and recent pedagogical material not present in commonly used corpora. This design minimizes overlap with publicly available graded datasets and reduces the possibility of memorization or contamination effects.
* All annotations undergo independent expert review and adjudication before inclusion in the benchmark. Neither the raw labels nor the precise scale definition are released publicly.
* The held-out scores, audio recordings, and annotations remain private to the task organizers throughout and after the evaluation. Participants submit executable systems (Docker containers preferred), which are run by the organizers on the hidden evaluation set.

Because the underlying difficulty scale is ordinal and intentionally hidden, system performance is evaluated primarily through rank correlation rather than exact class prediction. The official ranking metric is '''Kendall's Tau-c''', which measures agreement between predicted and reference difficulty orderings while remaining robust to unknown category spacing and differing numbers of ordinal levels.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is a single real-valued difficulty score. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

* '''July 1, 2026''': Submission system opens
* '''Oct 1, 2026''': Submission deadline
* '''Oct 15, 2026''': Results published

== Long-term Plan ==

We are committed to maintaining the task for at least three iterations. We will collaborate with the broader piano-difficulty research community to keep label-space conventions consistent across this task and parallel evaluation efforts. The hidden held-out test set is designed to grow incrementally across editions, with prior-year items optionally re-used subject to confirmation that participants have not gained access.

Given the field's relatively small size, we propose running this task annually for the first two years to bootstrap community engagement, and biennially thereafter. A symbolic-input track may be added in future editions if community interest warrants it.

== Organizers ==

* '''Pedro Ramoneda''' (Songscription) — [mailto:pedro@songscription.ai pedro@songscription.ai]
* '''Huan Zhang''' (Clefer) — [mailto:huan@clefer.com huan@clefer.com]

== References ==

* Ramoneda, P., Jeong, D., Eremenko, V., Tamer, N. C., Miron, M., & Serra, X. (2024). Combining piano performance dimensions for score difficulty classification. ''Expert Systems with Applications'', 238, 1–16.
* Ramoneda, P., Lee, M., Jeong, D., Valero-Mas, J. J., & Serra, X. (2025). Can audio reveal music performance difficulty? Insights from the Piano Syllabus Dataset.
* Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X., & Miron, M. (2022). Score difficulty analysis for piano performance education based on fingering. In ''ICASSP 2022 – IEEE International Conference on Acoustics, Speech and Signal Processing'' (pp. 201–205). IEEE.

2026:Music Performance Difficulty Prediction

2026-05-31T13:57:24Z

Huanz: /* Organizers */

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image-based, and audio-based modalities (Ramoneda et al., 2022–2025). Among these directions, audio-based difficulty estimation remains the most recent and least explored. Unlike symbolic or score-based approaches, audio-only methods do not require machine-readable scores, substantially widening the applicability of difficulty-prediction systems for music libraries, educational platforms, teachers, students, and recommendation systems.

Despite this growing interest, the field still lacks a standardized community benchmark for audio-based difficulty prediction. Existing studies rely on heterogeneous datasets, incompatible grading systems, and non-comparable evaluation protocols. This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty score.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': a real-valued difficulty score per recording. The score is treated as an ordering — its absolute scale, range, or number of categories does not need to match the (hidden) evaluation scale.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized score rendering from a fixed soundfont — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

There is no inference-time limit per piece, but the full test set must be processed within a 24-hour wall-clock budget on a single GPU.

== Significance of the Task ==

This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio, featuring:

* A consistent ordinal evaluation framework
* A fully private held-out test set curated specifically for this task
* Standardized ordinal regression and ranking metrics
* Composer-disjoint evaluation splits designed to measure true generalization beyond repertoire memorization
* A hidden pedagogical difficulty scale validated by expert pedagogues and musicologists

Beyond benchmarking, the task aims to lower the entry barrier for researchers working on audio representation learning, multimodal learning, and music understanding, while fostering a growing research community at the intersection of MIR, pedagogy, and computational music education.

== Evaluation Criteria ==

During development, participants are encouraged to evaluate their systems on public datasets such as PSyllabus using standard ordinal-regression metrics, including:

* Mean Squared Error (MSE)
* Accuracy within one level (Acc±1)
* Balanced accuracy
* Spearman's ρ

These metrics are useful for model selection and comparison on publicly annotated datasets with explicit grade structures.

However, because the official evaluation set uses a '''hidden ordinal pedagogical scale''' whose granularity and number of levels are not disclosed, the shared-task ranking is based exclusively on '''Kendall's Tau-c''', which measures agreement between predicted and reference orderings without assuming fixed distances between categories.

In addition to the official ranking metric, organizers may report supplementary analyses such as per-composer performance to better understand cross-repertoire generalization and robustness across stylistic domains.

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/12783403 PSyllabus]''' (Ramoneda et al., 2025): 7,901 audio recordings of classical piano pieces with 11-level pedagogical grades, together with mappings to multiple international grading systems. We specifically encourage participants to leverage the multiple ranking annotations available in PSyllabus (13 grading/ranking systems in total), not only the default labels, in order to study and improve generalization across heterogeneous pedagogical traditions and difficulty scales.
* '''[https://zenodo.org/records/8037327 CIPI (Can I Play It?)]''' (Ramoneda et al., 2024): 652 MusicXML piano pieces with 9-level Henle annotations, useful for participants interested in transcription-then-classification pipelines or distant supervision approaches.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''' (Ramoneda et al., 2022) and other publicly available piano difficulty datasets and pedagogical collections.

Participants may use any additional external data, provided that no part of the held-out evaluation repertoire is used directly or indirectly for training. To minimize contamination risks, a list of forbidden test pieces and composers will be published together with the task call.

Please describe in your technical report:
* Dataset names and sources
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To address the central concern of label trust and contamination, the held-out test set is constructed as a '''fully private benchmark''' curated specifically for this task. The dataset is assembled and annotated by expert piano pedagogues, with all annotations independently validated by a pedagogue and musicologist expert. Difficulty labels are assigned using an internal ordinal scale inspired by established pedagogical curricula and examination systems, but the exact mapping, granularity, and total number of levels are intentionally undisclosed to participants.

* The repertoire is selected from sources outside existing public piano difficulty datasets, including contemporary works, underrepresented composers, and recent pedagogical material not present in commonly used corpora. This design minimizes overlap with publicly available graded datasets and reduces the possibility of memorization or contamination effects.
* All annotations undergo independent expert review and adjudication before inclusion in the benchmark. Neither the raw labels nor the precise scale definition are released publicly.
* The held-out scores, audio recordings, and annotations remain private to the task organizers throughout and after the evaluation. Participants submit executable systems (Docker containers preferred), which are run by the organizers on the hidden evaluation set.

Because the underlying difficulty scale is ordinal and intentionally hidden, system performance is evaluated primarily through rank correlation rather than exact class prediction. The official ranking metric is '''Kendall's Tau-c''', which measures agreement between predicted and reference difficulty orderings while remaining robust to unknown category spacing and differing numbers of ordinal levels.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is a single real-valued difficulty score. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

(To be confirmed in line with the overall MIREX 2026 schedule.)

* '''TBD''': Task call published, training-data mapping released
* '''TBD''': Submission system opens
* '''TBD''': Submission deadline
* '''TBD''': Results announced at ISMIR 2026

== Long-term Plan ==

We are committed to maintaining the task for at least three iterations. We will collaborate with the broader piano-difficulty research community to keep label-space conventions consistent across this task and parallel evaluation efforts. The hidden held-out test set is designed to grow incrementally across editions, with prior-year items optionally re-used subject to confirmation that participants have not gained access.

Given the field's relatively small size, we propose running this task annually for the first two years to bootstrap community engagement, and biennially thereafter. A symbolic-input track may be added in future editions if community interest warrants it.

== Organizers ==

* '''Pedro Ramoneda''' (Songscription) — [mailto:pedro@songscription.ai pedro@songscription.ai]
* '''Huan Zhang''' (Clefer) — [mailto:huan@clefer.com huan@clefer.com]

== References ==

* Ramoneda, P., Jeong, D., Eremenko, V., Tamer, N. C., Miron, M., & Serra, X. (2024). Combining piano performance dimensions for score difficulty classification. ''Expert Systems with Applications'', 238, 1–16.
* Ramoneda, P., Lee, M., Jeong, D., Valero-Mas, J. J., & Serra, X. (2025). Can audio reveal music performance difficulty? Insights from the Piano Syllabus Dataset.
* Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X., & Miron, M. (2022). Score difficulty analysis for piano performance education based on fingering. In ''ICASSP 2022 – IEEE International Conference on Acoustics, Speech and Signal Processing'' (pp. 201–205). IEEE.

2026:Music Performance Difficulty Prediction

2026-05-31T13:57:09Z

Huanz: /* Organizers */

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image-based, and audio-based modalities (Ramoneda et al., 2022–2025). Among these directions, audio-based difficulty estimation remains the most recent and least explored. Unlike symbolic or score-based approaches, audio-only methods do not require machine-readable scores, substantially widening the applicability of difficulty-prediction systems for music libraries, educational platforms, teachers, students, and recommendation systems.

Despite this growing interest, the field still lacks a standardized community benchmark for audio-based difficulty prediction. Existing studies rely on heterogeneous datasets, incompatible grading systems, and non-comparable evaluation protocols. This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty score.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': a real-valued difficulty score per recording. The score is treated as an ordering — its absolute scale, range, or number of categories does not need to match the (hidden) evaluation scale.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized score rendering from a fixed soundfont — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

There is no inference-time limit per piece, but the full test set must be processed within a 24-hour wall-clock budget on a single GPU.

== Significance of the Task ==

This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio, featuring:

* A consistent ordinal evaluation framework
* A fully private held-out test set curated specifically for this task
* Standardized ordinal regression and ranking metrics
* Composer-disjoint evaluation splits designed to measure true generalization beyond repertoire memorization
* A hidden pedagogical difficulty scale validated by expert pedagogues and musicologists

Beyond benchmarking, the task aims to lower the entry barrier for researchers working on audio representation learning, multimodal learning, and music understanding, while fostering a growing research community at the intersection of MIR, pedagogy, and computational music education.

== Evaluation Criteria ==

During development, participants are encouraged to evaluate their systems on public datasets such as PSyllabus using standard ordinal-regression metrics, including:

* Mean Squared Error (MSE)
* Accuracy within one level (Acc±1)
* Balanced accuracy
* Spearman's ρ

These metrics are useful for model selection and comparison on publicly annotated datasets with explicit grade structures.

However, because the official evaluation set uses a '''hidden ordinal pedagogical scale''' whose granularity and number of levels are not disclosed, the shared-task ranking is based exclusively on '''Kendall's Tau-c''', which measures agreement between predicted and reference orderings without assuming fixed distances between categories.

In addition to the official ranking metric, organizers may report supplementary analyses such as per-composer performance to better understand cross-repertoire generalization and robustness across stylistic domains.

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/12783403 PSyllabus]''' (Ramoneda et al., 2025): 7,901 audio recordings of classical piano pieces with 11-level pedagogical grades, together with mappings to multiple international grading systems. We specifically encourage participants to leverage the multiple ranking annotations available in PSyllabus (13 grading/ranking systems in total), not only the default labels, in order to study and improve generalization across heterogeneous pedagogical traditions and difficulty scales.
* '''[https://zenodo.org/records/8037327 CIPI (Can I Play It?)]''' (Ramoneda et al., 2024): 652 MusicXML piano pieces with 9-level Henle annotations, useful for participants interested in transcription-then-classification pipelines or distant supervision approaches.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''' (Ramoneda et al., 2022) and other publicly available piano difficulty datasets and pedagogical collections.

Participants may use any additional external data, provided that no part of the held-out evaluation repertoire is used directly or indirectly for training. To minimize contamination risks, a list of forbidden test pieces and composers will be published together with the task call.

Please describe in your technical report:
* Dataset names and sources
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To address the central concern of label trust and contamination, the held-out test set is constructed as a '''fully private benchmark''' curated specifically for this task. The dataset is assembled and annotated by expert piano pedagogues, with all annotations independently validated by a pedagogue and musicologist expert. Difficulty labels are assigned using an internal ordinal scale inspired by established pedagogical curricula and examination systems, but the exact mapping, granularity, and total number of levels are intentionally undisclosed to participants.

* The repertoire is selected from sources outside existing public piano difficulty datasets, including contemporary works, underrepresented composers, and recent pedagogical material not present in commonly used corpora. This design minimizes overlap with publicly available graded datasets and reduces the possibility of memorization or contamination effects.
* All annotations undergo independent expert review and adjudication before inclusion in the benchmark. Neither the raw labels nor the precise scale definition are released publicly.
* The held-out scores, audio recordings, and annotations remain private to the task organizers throughout and after the evaluation. Participants submit executable systems (Docker containers preferred), which are run by the organizers on the hidden evaluation set.

Because the underlying difficulty scale is ordinal and intentionally hidden, system performance is evaluated primarily through rank correlation rather than exact class prediction. The official ranking metric is '''Kendall's Tau-c''', which measures agreement between predicted and reference difficulty orderings while remaining robust to unknown category spacing and differing numbers of ordinal levels.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is a single real-valued difficulty score. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

(To be confirmed in line with the overall MIREX 2026 schedule.)

* '''TBD''': Task call published, training-data mapping released
* '''TBD''': Submission system opens
* '''TBD''': Submission deadline
* '''TBD''': Results announced at ISMIR 2026

== Long-term Plan ==

We are committed to maintaining the task for at least three iterations. We will collaborate with the broader piano-difficulty research community to keep label-space conventions consistent across this task and parallel evaluation efforts. The hidden held-out test set is designed to grow incrementally across editions, with prior-year items optionally re-used subject to confirmation that participants have not gained access.

Given the field's relatively small size, we propose running this task annually for the first two years to bootstrap community engagement, and biennially thereafter. A symbolic-input track may be added in future editions if community interest warrants it.

== Organizers ==

* '''Pedro Ramoneda''' (Songscription) — [mailto:pedro@songscription.ai pedro@songscription.ai]
* '''Huan Zhang''' (Clefer) — [mailto:huan@clefer.com huan.zhang@qmul.ac.uk]

== References ==

* Ramoneda, P., Jeong, D., Eremenko, V., Tamer, N. C., Miron, M., & Serra, X. (2024). Combining piano performance dimensions for score difficulty classification. ''Expert Systems with Applications'', 238, 1–16.
* Ramoneda, P., Lee, M., Jeong, D., Valero-Mas, J. J., & Serra, X. (2025). Can audio reveal music performance difficulty? Insights from the Piano Syllabus Dataset.
* Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X., & Miron, M. (2022). Score difficulty analysis for piano performance education based on fingering. In ''ICASSP 2022 – IEEE International Conference on Acoustics, Speech and Signal Processing'' (pp. 201–205). IEEE.

2026:Music Performance Difficulty Prediction

2026-05-30T23:06:56Z

Huanz: /* Music Performance Difficulty Prediction 2026 */

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image-based, and audio-based modalities (Ramoneda et al., 2022–2025). Among these directions, audio-based difficulty estimation remains the most recent and least explored. Unlike symbolic or score-based approaches, audio-only methods do not require machine-readable scores, substantially widening the applicability of difficulty-prediction systems for music libraries, educational platforms, teachers, students, and recommendation systems.

Despite this growing interest, the field still lacks a standardized community benchmark for audio-based difficulty prediction. Existing studies rely on heterogeneous datasets, incompatible grading systems, and non-comparable evaluation protocols. This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty score.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': a real-valued difficulty score per recording. The score is treated as an ordering — its absolute scale, range, or number of categories does not need to match the (hidden) evaluation scale.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized score rendering from a fixed soundfont — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

There is no inference-time limit per piece, but the full test set must be processed within a 24-hour wall-clock budget on a single GPU.

== Significance of the Task ==

This shared task establishes the first community-wide benchmark for piano difficulty estimation from audio, featuring:

* A consistent ordinal evaluation framework
* A fully private held-out test set curated specifically for this task
* Standardized ordinal regression and ranking metrics
* Composer-disjoint evaluation splits designed to measure true generalization beyond repertoire memorization
* A hidden pedagogical difficulty scale validated by expert pedagogues and musicologists

Beyond benchmarking, the task aims to lower the entry barrier for researchers working on audio representation learning, multimodal learning, and music understanding, while fostering a growing research community at the intersection of MIR, pedagogy, and computational music education.

== Evaluation Criteria ==

During development, participants are encouraged to evaluate their systems on public datasets such as PSyllabus using standard ordinal-regression metrics, including:

* Mean Squared Error (MSE)
* Accuracy within one level (Acc±1)
* Balanced accuracy
* Spearman's ρ

These metrics are useful for model selection and comparison on publicly annotated datasets with explicit grade structures.

However, because the official evaluation set uses a '''hidden ordinal pedagogical scale''' whose granularity and number of levels are not disclosed, the shared-task ranking is based exclusively on '''Kendall's Tau-c''', which measures agreement between predicted and reference orderings without assuming fixed distances between categories.

In addition to the official ranking metric, organizers may report supplementary analyses such as per-composer performance to better understand cross-repertoire generalization and robustness across stylistic domains.

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/12783403 PSyllabus]''' (Ramoneda et al., 2025): 7,901 audio recordings of classical piano pieces with 11-level pedagogical grades, together with mappings to multiple international grading systems. We specifically encourage participants to leverage the multiple ranking annotations available in PSyllabus (13 grading/ranking systems in total), not only the default labels, in order to study and improve generalization across heterogeneous pedagogical traditions and difficulty scales.
* '''[https://zenodo.org/records/8037327 CIPI (Can I Play It?)]''' (Ramoneda et al., 2024): 652 MusicXML piano pieces with 9-level Henle annotations, useful for participants interested in transcription-then-classification pipelines or distant supervision approaches.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''' (Ramoneda et al., 2022) and other publicly available piano difficulty datasets and pedagogical collections.

Participants may use any additional external data, provided that no part of the held-out evaluation repertoire is used directly or indirectly for training. To minimize contamination risks, a list of forbidden test pieces and composers will be published together with the task call.

Please describe in your technical report:
* Dataset names and sources
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To address the central concern of label trust and contamination, the held-out test set is constructed as a '''fully private benchmark''' curated specifically for this task. The dataset is assembled and annotated by expert piano pedagogues, with all annotations independently validated by a pedagogue and musicologist expert. Difficulty labels are assigned using an internal ordinal scale inspired by established pedagogical curricula and examination systems, but the exact mapping, granularity, and total number of levels are intentionally undisclosed to participants.

* The repertoire is selected from sources outside existing public piano difficulty datasets, including contemporary works, underrepresented composers, and recent pedagogical material not present in commonly used corpora. This design minimizes overlap with publicly available graded datasets and reduces the possibility of memorization or contamination effects.
* All annotations undergo independent expert review and adjudication before inclusion in the benchmark. Neither the raw labels nor the precise scale definition are released publicly.
* The held-out scores, audio recordings, and annotations remain private to the task organizers throughout and after the evaluation. Participants submit executable systems (Docker containers preferred), which are run by the organizers on the hidden evaluation set.

Because the underlying difficulty scale is ordinal and intentionally hidden, system performance is evaluated primarily through rank correlation rather than exact class prediction. The official ranking metric is '''Kendall's Tau-c''', which measures agreement between predicted and reference difficulty orderings while remaining robust to unknown category spacing and differing numbers of ordinal levels.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is a single real-valued difficulty score. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

(To be confirmed in line with the overall MIREX 2026 schedule.)

* '''TBD''': Task call published, training-data mapping released
* '''TBD''': Submission system opens
* '''TBD''': Submission deadline
* '''TBD''': Results announced at ISMIR 2026

== Long-term Plan ==

We are committed to maintaining the task for at least three iterations. We will collaborate with the broader piano-difficulty research community to keep label-space conventions consistent across this task and parallel evaluation efforts. The hidden held-out test set is designed to grow incrementally across editions, with prior-year items optionally re-used subject to confirmation that participants have not gained access.

Given the field's relatively small size, we propose running this task annually for the first two years to bootstrap community engagement, and biennially thereafter. A symbolic-input track may be added in future editions if community interest warrants it.

== Organizers ==

* '''Huan Zhang''' (Queen Mary University of London) — [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]
* '''Pedro Ramoneda''' (Songscription) — [mailto:pedro@songscription.ai pedro@songscription.ai]

== References ==

* Ramoneda, P., Jeong, D., Eremenko, V., Tamer, N. C., Miron, M., & Serra, X. (2024). Combining piano performance dimensions for score difficulty classification. ''Expert Systems with Applications'', 238, 1–16.
* Ramoneda, P., Lee, M., Jeong, D., Valero-Mas, J. J., & Serra, X. (2025). Can audio reveal music performance difficulty? Insights from the Piano Syllabus Dataset.
* Ramoneda, P., Tamer, N. C., Eremenko, V., Serra, X., & Miron, M. (2022). Score difficulty analysis for piano performance education based on fingering. In ''ICASSP 2022 – IEEE International Conference on Acoustics, Speech and Signal Processing'' (pp. 201–205). IEEE.

2026:Music Performance Difficulty Prediction

2026-05-30T14:22:20Z

Huanz: Created page with "= Music Performance Difficulty Prediction 2026 = Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This..."

= Music Performance Difficulty Prediction 2026 =

Welcome to the official MIREX wiki page for the '''Music Performance Difficulty Prediction''' task, new for MIREX 2026. This task targets the automatic estimation of how technically and musically demanding a piece is to perform from audio recordings, supporting applications in music education, library cataloguing, and pedagogical recommendation systems.

== Overview ==

Music performance difficulty prediction has emerged as an active MIR sub-field over the past five years, with foundational datasets (Mikrokosmos-difficulty, CIPI, PSyllabus) and methods spanning symbolic, image, and audio modalities ([https://arxiv.org/abs/2203.13010 Ramoneda et al., ICASSP 2022]; [https://arxiv.org/abs/2306.08480 2023]; [https://arxiv.org/abs/2309.16287 ISMIR 2023]; [https://arxiv.org/abs/2403.03947 TASLP 2024]; [https://arxiv.org/abs/2408.00473 ISMIR 2024]). Despite this growth, there has been no community-wide evaluation: published results use inconsistent splits, label scales, and metrics, making cross-method comparison difficult.

This task establishes the first shared benchmark for audio-based music performance difficulty prediction, with a consistent ordinal label space, a held-out test set never publicly released, and standard ordinal regression metrics under composer-disjoint splits.

== Task Description ==

Participants submit systems that take a '''solo piano audio recording''' as input and output a predicted difficulty grade on an ordinal '''1–9 scale''', aligned to the Henle Verlag grading system.

* '''Input format''': WAV, 44.1 kHz, mono or stereo.
* '''Output format''': integer grade 1–9 together with a 9-dimensional confidence vector.
* '''Two recordings per piece''': each test piece is represented by at least two recordings — one human performance and one synthesized rendering — scored independently and aggregated per piece. This is designed so that systems cannot game the task by relying purely on performance-quality or recording-condition cues.

This year we deliberately focus on '''audio only''' to align with the broader applicability of audio-based difficulty prediction tools (no machine-readable score required) and to attract participants from the audio representation learning community. A symbolic-input track may be added in future editions.

== Evaluation Criteria ==

Submissions are ranked using standard ordinal regression metrics computed on the held-out test set:

* '''Mean Squared Error (MSE)''' — primary ranking metric
* '''Accuracy within one level (Acc±1)''' — fraction of predictions within one grade of the true label
* '''Balanced accuracy''' — per-class accuracy averaged across grade levels, to handle class imbalance
* '''Spearman ρ''' — rank correlation between predicted and ground-truth grades

The primary ranking uses MSE; ties are broken by Acc±1. We additionally report per-difficulty-level confusion matrices, per-composer error, and the gap between human-recording and synthesized-recording scores on the same piece (to surface systems that overfit to performance-quality cues).

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we require full transparency in the technical report.

Suggested datasets for training and validation:

* '''[https://zenodo.org/records/12783403 PSyllabus]''' (Ramoneda et al., 2024): 7,901 audio recordings of classical piano pieces with 11-level grades. A provided mapping to the Henle 1–9 label space will be released alongside the task.
* '''[https://zenodo.org/records/8037327 CIPI (Can I Play It?)]''' (Ramoneda et al., 2023): 652 MusicXML pieces with 9-level Henle labels. Useful for participants who wish to transcribe-then-classify or for distant supervision.
* '''[https://github.com/PRamoneda/Mikrokosmos-difficulty Mikrokosmos-difficulty]''': 147 Bartók pieces with 3 difficulty levels.
* Other publicly available difficulty collections (Pianostreet-difficulty, Freescore-difficulty, Hidden Voices) may also be used.

'''Important''': no part of the held-out test set may be used for training. A list of forbidden pieces will be published with the task call. Please describe in your technical report:
* Dataset name and source
* Size and number of pieces
* Any preprocessing, cleaning, or label-space remapping applied

== Held-Out Test Set ==

To ensure label trust and prevent training-set contamination, the held-out test set (target size '''≈ 200 pieces''') is constructed by the task captain in three stages and is '''not publicly released''':

# '''Piece selection''': drawn from sources outside existing public difficulty datasets — (a) Henle Verlag catalogue additions published after the CIPI cutoff; (b) ABRSM, Trinity, and RCM 2024–2026 syllabus pieces not present in PSyllabus; (c) a curated set of contemporary and historically underrepresented composers (e.g. Hidden Voices) graded specifically for this task.
# '''Grade verification''': each piece is graded independently by at least two expert pianists recruited from the QMUL music programme and the ABRSM examiner network on the Henle 1–9 scale. Inter-annotator agreement is reported alongside the final results. Disagreements >1 level are resolved by a third annotator.
# '''Audio rendering''': for each piece, at least two reference recordings are produced — (i) a human performance, either recorded on a Yamaha Disklavier at QMUL or sourced from licensed commercial recordings where redistribution of features is permitted; and (ii) a high-quality synthesized rendering from the verified score, generated with a fixed renderer.

'''Test set governance''': the held-out audio files and labels remain private to the task captain. Participants submit systems; the task captain runs inference on the held-out set and reports metrics back. Test recordings and labels are not released after the competition, enabling re-use and incremental growth across editions.

== Submission Requirements ==

The following items are required for submission:

* '''System''': packaged as a Docker container with a standardised inference interface — input is a path to a WAV file; output is an integer grade 1–9 with a 9-dimensional confidence vector. A reference Docker template and inference wrapper script will be provided.
* '''Technical report''': 2–4 pages in ISMIR LBD format describing training data, model architecture, label-space handling, and any post-processing.
* '''Compute declaration''': GPU memory footprint and average inference time per piece.

Final submission must be made through the [[2026_Submission_Guidelines|MIREX submission system]]. Submissions exceeding the 24-hour total inference budget on the full test set, or failing on more than 5% of test items, are reported but excluded from the primary ranking.

== Timeline ==

(To be confirmed in line with the overall MIREX 2026 schedule.)

* '''TBD''': Task call published, training-data mapping released
* '''TBD''': Submission system opens
* '''TBD''': Submission deadline
* '''TBD''': Results announced at ISMIR 2026

== Organizers ==

* '''Huan Zhang''' (Task Captain, Queen Mary University of London)
* '''Pedro Ramoneda''' (Universitat Pompeu Fabra)

Contact: [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]

2025:RenCon Results

2025-09-27T15:22:12Z

Huanz:

= 2025:RenCon Results =

== Preliminary (Audition) Round Results ==

=== Evaluation Methodology ===
The preliminary round was evaluated through an online listening test with '''25 expert evaluators'''. The evaluation used a weighted voting system where participants self-rated their expertise level from 1-5 stars, with responses weighted accordingly.

=== Participant Demographics ===
Our evaluation panel consisted of highly qualified judges:

'''Expertise Distribution:'''
* Expert evaluators (5 stars): 7 participants (29.2%)
* High confidence (4 stars): 5 participants (20.8%)
* Moderate confidence (3 stars): 10 participants (41.7%)
* Lower confidence (1-2 stars): 2 participants (8.4%)
* '''Average expertise weight:''' 3.67/5.0

'''Professional Background:'''
* Music researchers: 12 (54.5%)
* Music technologists: 10 (45.5%)
* Active performers: 8 (36.4%)
* Conservatory students: 6 (27.3%)
* Music lovers: 15 (68.2%)
* Concert-goers: 8 (36.4%)

'''Musical Experience:'''
* Strong representation of classical music expertise
* Diverse musical preferences spanning classical, jazz, pop, and rock
* Substantial piano experience among evaluators
* Mix of academic researchers and practicing musicians

=== System Rankings ===

The following table shows the final rankings based on weighted average scores from the preliminary round evaluation:
{| class="wikitable sortable"
! Rank
! Anonymous Name
! Real System Name
! Authors/Institution
! Weighted Score
! PDF
|-
| 1
| MidnightOpal
| DirectorMusices
| Anders Friberg, Gabriel Jones
| 4.33/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/DirectorMusices.pdf PDF]
|-
| 2
| CrystalEcho
| VirtuosoNet
| Dasaem Jeong, Taegyun Kwon, Yoojin Kim, Kyogu Lee, Juhan Nam
| 3.54/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/VirtuosoNet.pdf PDF]
|-
| 3
| FrozenRiver
| Midihum
| Erich Grunewald
| 3.32/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/Midihum.pdf PDF]
|-
| 4
| VelvetStorm
| ElegantAIPianist
| Leduo Chen, Xinrui Su, Yuqiang Li, Honyu Andy Shing, Junchuan Zhao, Zihan Chai, Kunyang Zhang, Shengchen Li
| 3.19/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/ElegantAIPianist.pdf PDF]
|-
| 5
| SilverWave
| Contin-U
| Jongmin Jung, Dongmin Kim, Sihun Lee, Seola Cho, Hyungjoon Soh, Irmak Bukey, Chris Donahue, Dasaem Jeong
| 3.00/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/Contin-U.pdf PDF]
|-
| 6
| EmberSky
| YQX+
| Jinwen Zhou, Yuncong Xie, Haochen Wang, Huan Zhang, Aidan Hogg, Simon Dixon
| 2.83/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/YQX+.pdf PDF]
|-
|
| CrimsonDawn
| ScorePerLockNAR
| Weixi Zhai
| 2.53/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/ScorePerLockNAR.pdf PDF]
|-
|
| AzureThunder
| RenConnoisseur
| Silvan Peter
| 2.53/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/RenConnoisseur.pdf PDF]
|-
|
| GoldenMist
| CueFreeExpressPedal
| Kyle Worrall, Tom Collins
| 2.31/5.0
| [https://futuremirex.com/portal/wp-content/uploads/2025/rencon/CueFreeExpressPedal.pdf PDF]
|}

''Note: Complete rankings and system details will be updated following the live contest and final results announcement.''

=== Qualitative Feedback ===

Evaluators provided extensive qualitative feedback on the systems' performances:

'''Common Positive Attributes:'''
* Natural expressiveness and human-like phrasing
* Appropriate tempo variations and rubato
* Musical sensitivity to harmonic structure
* Dynamic expression and articulation

'''Areas for Improvement:'''
* Consistency across different musical styles
* Handling of complex rhythmic patterns
* Balance between technical accuracy and musical expression

== Live Contest Results ==

The live contest was held on September 25, 2025, at the Daejeon Convention Center during ISMIR 2025. All systems rendered a surprise piece in real-time, with audience voting determining the final rankings. A total of '''50 evaluators''' participated in the live evaluation.

=== Surprise Piece ===
* '''Title:''' Variation on "Mother and Sister"
* '''Composer:''' Hayeon Bang
* '''Duration:''' ~3 minutes

=== Live Contest Rankings ===

{| class="wikitable sortable"
! Rank
! System Index
! Model Name
! Authors
! Live Score
! Preliminary Rank
! Score Change
|-
| 1st
| #8
| VirtuosoNet
| Dasaem Jeong et al.
| 3.62/5.0
| 2nd
| ↑1
|-
| 2nd
| #1
| DirectorMusices
| Anders Friberg et al.
| 3.06/5.0
| 1st
| ↓1
|-
| 3rd
| #2
| Midihum
| Erich Grunewald
| 2.90/5.0
| 3rd
| —
|-
| 4th
| #4
| Contin-U
| Jongmin Jung et al.
| 2.90/5.0
| 5th
| ↑1
|-
| 5th
| #6
| ScorePerLockNAR
| Weixi Zhai
| 2.52/5.0
| 7th
| ↑2
|-
| 6th
| #3
| RenConnoisseur
| Silvan Peter
| 2.40/5.0
| 8th
| ↑2
|-
| 7th
| #9
| ElegantAIPianist
| Leduo Chen et al.
| 2.08/5.0
| 4th
| ↓3
|-
| 8th
| #5
| YQX+
| Jinwen Zhou et al.
| 1.79/5.0
| 6th
| ↓2
|}

=== Human Baseline Performance ===
A human performance of the same surprise piece was also evaluated by the audience as a reference point:

{| class="wikitable"
! Performance Type
! Score
! Rank Among All Performances
|-
| Human Performance
| 4.40/5.0
| 1st (Overall)
|}

=== Performance Analysis ===

'''Top Performers:'''
* '''VirtuosoNet''' achieved the highest AI system score (3.62/5.0), improving from 2nd place in the preliminary round
* '''DirectorMusices''' maintained strong performance (3.06/5.0) but dropped from 1st to 2nd place
* '''Midihum''' showed consistency, maintaining 3rd place across both rounds

'''Notable Changes:'''
* '''ScorePerLockNAR''' and '''RenConnoisseur''' both improved by 2 positions in the live contest
* '''ElegantAIPianist''' experienced the largest drop, falling from 4th to 7th place
* '''Contin-U''' improved from 5th to 4th place in live evaluation

'''Human vs. AI Performance:'''
* The human performance scored 4.40/5.0, significantly higher than the top AI system
* This represents a performance gap of 0.78 points between human and best AI performance
* All AI systems scored below the human baseline, indicating continued room for improvement

=== Final Winner ===

'''🏆 RenCon 2025 Winner: VirtuosoNet'''
* '''Team:''' Dasaem Jeong, Taegyun Kwon, Yoojin Kim, Kyogu Lee, Juhan Nam
* '''Final Score:''' 3.62/5.0
* '''Achievement:''' Highest-scoring AI system in live contest evaluation

The winner was announced at the closing ceremony of ISMIR 2025 on September 25, 2025.

== External Links ==

* [https://ren-con2025.vercel.app/ Official RenCon 2025 Website]
* [https://ismir2025.ismir.net/ ISMIR 2025 Conference Website]
* [https://www.music-ir.org/mirex/wiki/2025:RenCon RenCon 2025 MIREX Task Page]

[[Category:MIREX]]
[[Category:ISMIR 2025]]
[[Category:Performance Rendering]]
[[Category:Competition Results]]

2025:RenCon Results

2025-09-18T23:29:30Z

Huanz:

2025:RenCon Results

2025-09-18T23:21:18Z

Huanz:

2025:RenCon Results

2025-09-11T08:41:09Z

Huanz: Blanked the page

2025:RenCon Results

2025-09-11T08:41:01Z

Huanz: /* Next Steps */

2025:RenCon Results

2025-09-11T08:40:55Z

Huanz: /* About the Scoring */

# RenCon 2025 Online Audition Results - Preliminary

'''Note: These are preliminary results as of September 10, 2025. The deadline has not yet passed, and rankings may change as additional responses are received.'''

== Next Steps ==

* '''Submission deadline''': Still open for additional audience feedback
* '''Final results''': Will be announced after the deadline closes
* '''Advancement''': Top performers will proceed to the next round

''Last updated: September 10, 2025''

2025:RenCon Results

2025-09-11T08:19:41Z

Huanz: /* Current Rankings */

# RenCon 2025 Online Audition Results - Preliminary

'''Note: These are preliminary results as of September 10, 2025. The deadline has not yet passed, and rankings may change as additional responses are received.'''

== About the Scoring ==

* Scores are calculated using a weighted average system where judges self-rate their expertise level (1-5 stars)
* Higher expertise ratings receive greater weight in the final calculations
* All 9 participants received ratings from all 13 current respondents

== Next Steps ==

* '''Submission deadline''': Still open for additional audience feedback
* '''Final results''': Will be announced after the deadline closes
* '''Advancement''': Top performers will proceed to the next round

''Last updated: September 10, 2025''

2025:RenCon Results

2025-09-10T23:44:48Z

Huanz: Created page with "# RenCon 2025 Online Audition Results - Preliminary '''Note: These are preliminary results as of September 10, 2025. The deadline has not yet passed, and rankings may change..."

# RenCon 2025 Online Audition Results - Preliminary

'''Note: These are preliminary results as of September 10, 2025. The deadline has not yet passed, and rankings may change as additional responses are received.'''

== Current Rankings ==

Based on weighted audience ratings from '''13 survey responses''':

{| class="wikitable"
|+ Audition Results
! Rank !! Performer !! Weighted Score !! Status
|-
| 1st || '''MidnightOpal''' || 4.21/5.0 || Leading
|-
| 2nd || '''CrystalEcho''' || 3.56/5.0 || Strong showing
|-
| 3rd || '''VelvetStorm''' || 3.29/5.0 || Competitive
|-
| 4th || '''SilverWave''' || 3.19/5.0 || Solid performance
|-
| 5th || '''FrozenRiver''' || 3.17/5.0 || Close contest
|-
| 6th || '''EmberSky''' || 2.67/5.0 ||
|-
| 7th || '''AzureThunder''' || 2.50/5.0 ||
|-
| 8th || '''CrimsonDawn''' || 2.40/5.0 ||
|-
| 9th || '''GoldenMist''' || 2.06/5.0 ||
|}

== About the Scoring ==

* Scores are calculated using a weighted average system where judges self-rate their expertise level (1-5 stars)
* Higher expertise ratings receive greater weight in the final calculations
* All 9 participants received ratings from all 13 current respondents

== Next Steps ==

* '''Submission deadline''': Still open for additional audience feedback
* '''Final results''': Will be announced after the deadline closes
* '''Advancement''': Top performers will proceed to the next round

''Last updated: September 10, 2025''

2025:RenCon

2025-08-19T22:18:58Z

Huanz: /* Timeline */

= RenCon 2025: Expressive Performance Rendering Competition =

Welcome to the official MIREX wiki page for RenCon 2025, a new task on expressive musical rendering from symbolic scores. This task is part of the MIREX 2025 evaluation campaign and will culminate with a live contest at ISMIR 2025 in Daejeon, Korea. For updates, templates, and examples, please visit our official site: https://ren-con2025.vercel.app

Here is a short summary of the important information:

== Overview ==

Welcome to the official site for '''RenCon 2025''', an international challenge where researchers and developers submit systems capable of rendering expressive musical performances from symbolic scores. This year, we are delighted to hosted it along with '''ISMIR 2025''' and under the '''MIREX Tasks protocol'''.

The RenCon competition has a rich history, having record for compeition in 2002, 2003, 2004, 2005, 2008 and 2011 ([https://www.researchgate.net/publication/228715822_The_Second_Rencon_Performance_Contest_Panel_Discussion_and_the_Future 1], [https://www.nime.org/proceedings/2004/nime2004_120.pdf 2], [https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=e611d7c4f9df63d99d5760ca88a7107dca945e05 3], [https://www.researchgate.net/publication/228715816_Rencon_Performance_Rendering_Contest_for_Automated_Music_Systems 4], [http://smc.afim-asso.org/smc11/papers/smc2011_120.pdf 5]), jointly with conferences like SMC, NIME, ICMPC and IJCAI. Even before the term "AI" was widely used, the RenCon competition has been a platform for researchers to showcase their work in the field of expressive performance rendering.
However, not too much of its past can be traced today, except an old [http://renconmusic.org site]. We hope to revive this tradition with RenCon 2025, coinciding with the renewed global focus on performance during this year's Chopin Piano Competition.

== Task Description ==

Expressive Performance Rendering is a task that challenges participants to develop systems capable of rendering expressive musical performances from symbolic scores in MusicXML format.

This year, we limit the task to be '''solo-piano''' specific. We accept system that generate symbolic (MIDI) or audio (wav) renderings, and the output shall contain human-like expressive deviation from the MusicXML score.

Similar to AI song contest, the evaluation of expressive rendering is subjective and requires human judges to assess the quality of the generated performances. Thus, we propose a two-phase competition structure as shown in the next section, relying on audience voting to determine the winner.

== Competition Structure ==

RenCon 2025 is structured in two phases:

* '''Phase 1 – Preliminary Round (Online)'''
Submit performances of assigned and free-choice pieces. Includes symbolic or audio renderings and a technical report. No real-time constraints allow for broader participation and diversity in evaluation.
The submission period is open from '''May 30, 2025''' to '''Aug 20, 2025''', following the MIREX submission guidelines.
After the submission deadline, the preliminary round page will be finalized with the list of participants and their submissions, and the online evaluation will take place.

* '''Phase 2 – Live Contest at ISMIR (Daejeon, Korea)'''
Top systems from preliminary round will be invited to render a surprise piece live at ISMIR, using their system in real time.
The live contest is open to all ISMIR attendees, as well as general public if the venue allows. The audience will be able to listen to the live performances and vote for their favorite system. The winner will be announced at the end of the conference.

== Timeline ==

* '''May 30, 2025''': Preliminary round submission opens, online judge sign-up opens
* '''Aug 25, 2025''': Preliminary round submission closes (extended deadline)
* '''Aug 25, 2025''': Preliminary round page finalized and online evaluation begins
* '''Sept 10, 2025''': Online evaluation ends, results announced, and top systems invited to live contest
* '''Sept 25, 2025''': Live contest at ISMIR 2025 in Daejeon, Korea

== Submission Requirements ==

The following items are required for submission:

* Code and checkpoint of the system.
* Symbolic (MIDI) or audio (wav) renderings of designated pieces:
** Required pieces (choose 2 out of 4), click to download the MusicXML files:
*** [https://ren-con2025.vercel.app/static/pieces/CAPRICCIO_en_sol_mineur_HWV_483_-_Handel.mxl Handel: Capriccio in G minor, HWV 483]
*** [https://ren-con2025.vercel.app/static/pieces/32_Variations_in_C_minor_WoO_80_First_5.mxl Beethoven: 32 Variations in C minor, WoO 80 - Theme and the first 5 variations]
*** [https://ren-con2025.vercel.app/static/pieces/12_Romances_Op.21__Sergei_Rachmaninoff_Zdes_khorosho_-_Arrangement_for_solo_piano.mxl Rachmaninoff: Здесь хорошо (How Fair this Spot), Op. 21, No. 7 (transcribed for solo piano)]
*** [https://ren-con2025.vercel.app/static/pieces/With_Dog-teams.mxl Amy Beach: Eskimos, Op.64, No.4 - With Dog-Teams]
** One free-choice piece (rendering must be less than 1 minute 30 seconds)
** A total of 3 pieces, with a maximum of 5 minutes of rendering in total.

* Technical report: Please use the template available at [https://ren-con2025.vercel.app/static/RenCon%20Submission%20Report%20Template.zip Submission Report Template (ZIP)]

Final submission must be made through the [[2025_Submission_Guidelines|MIREX submission system]], with maximum zip file of size 5GB.

== Data Format of submission ==

'''Symbolic submissions''': MIDI format with all sound events in program number 1 (solo piano). Track and channel are unrestricted.

'''Audio Submissions''': wav format, 44.1kHz, 16-bit PCM

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we ask for full transparency.

Some suggested datasets for training and validation include:
* '''[https://github.com/fosfrancesco/asap-dataset ASAP]''': A large dataset of classical piano performances sourced from MAESTRO, includes corresponding MIDI and audio. [https://github.com/CPJKU/asap-dataset (n)ASAP] provides score-performance alignment.
* '''[https://github.com/tangjjbetsy/ATEPP ATEPP]''': A large dataset of transcribed MIDI expressive piano performances, organized by virtuosic performer. However, only around half of the dataset contains score MusicXML files.
* '''[https://github.com/CPJKU/vienna4x22 VIENNA 4x22]''': A small-scale dataset of 4 pieces with 22 different interpretations, including audio and MIDI and fine alignment. ()
* '''[https://github.com/huispaty/batik_plays_mozart Batik-plays-Mozart]''': Fine-aligned performance MIDI dataset of Mozart played by Roland Batik.

Please clearly describe the datasets used for training and validation in your technical report. Important details to include are:

* Dataset name or source
* Size and number of pieces
* Instrumentation and expressive characteristics
* Data format (MIDI, audio, etc.)
* Any preprocessing, cleaning, or augmentation steps applied

This helps the jury and the research community understand the representational capacity and limitations of each submission.

== Post-Processing ==

To ensure fair evaluation, all post-processing applied to the preliminary round output must be documented in the submission report. Depending on your system type, please include the following:

* '''Symbolic Output System''': If your model generates symbolic MIDI output and you submit the sonified audio track, describe how audio is derived. Include soundfont names, software synths used (e.g., FluidSynth, Logic Pro), or player piano models.
** If you would like to submit the MIDI output directly and allow us (the organizer team) for sonification, please contact [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk] during your submission. It's likely that we would arrange a Disklavier recording in the Vienna office of Institute of Computational Perception (CPJKU) lab.
* '''Audio Output Systems''': If your model outputs audio directly, describe if you have applied any enhancement steps such as EQ, reverb, compression, or noise reduction to the model's output.
* '''Controllability or Interventions''': Clarify if the output is influenced by human-involved choices — such as selected tempo, dynamics range, segmentation, or annotated phrasing.
* '''MIDI Cleanup''': If symbolic outputs were manually edited (quantization, pedals, etc) before submission, that should be documented.

Submissions should aim for minimal human intervention. Manual correction is allowed only if it is well-documented and justified in the report.

== Organizers ==
* Huan Zhang (Task Captain, Queen Mary University of London)
* Taegyun Kwon, Venue Coordinator, Korea Advanced Institute of Science and Technology (KAIST)
* Junyan Jiang (New York University)
* Simon Dixon (Queen Mary University of London)
* Gus Xia (MBZUAI)
* Akira Maezawa (Yamaha)

Contact: [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]

2025:RenCon

2025-08-19T22:18:48Z

Huanz: /* Timeline */

= RenCon 2025: Expressive Performance Rendering Competition =

Welcome to the official MIREX wiki page for RenCon 2025, a new task on expressive musical rendering from symbolic scores. This task is part of the MIREX 2025 evaluation campaign and will culminate with a live contest at ISMIR 2025 in Daejeon, Korea. For updates, templates, and examples, please visit our official site: https://ren-con2025.vercel.app

Here is a short summary of the important information:

== Overview ==

Welcome to the official site for '''RenCon 2025''', an international challenge where researchers and developers submit systems capable of rendering expressive musical performances from symbolic scores. This year, we are delighted to hosted it along with '''ISMIR 2025''' and under the '''MIREX Tasks protocol'''.

The RenCon competition has a rich history, having record for compeition in 2002, 2003, 2004, 2005, 2008 and 2011 ([https://www.researchgate.net/publication/228715822_The_Second_Rencon_Performance_Contest_Panel_Discussion_and_the_Future 1], [https://www.nime.org/proceedings/2004/nime2004_120.pdf 2], [https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=e611d7c4f9df63d99d5760ca88a7107dca945e05 3], [https://www.researchgate.net/publication/228715816_Rencon_Performance_Rendering_Contest_for_Automated_Music_Systems 4], [http://smc.afim-asso.org/smc11/papers/smc2011_120.pdf 5]), jointly with conferences like SMC, NIME, ICMPC and IJCAI. Even before the term "AI" was widely used, the RenCon competition has been a platform for researchers to showcase their work in the field of expressive performance rendering.
However, not too much of its past can be traced today, except an old [http://renconmusic.org site]. We hope to revive this tradition with RenCon 2025, coinciding with the renewed global focus on performance during this year's Chopin Piano Competition.

== Task Description ==

Expressive Performance Rendering is a task that challenges participants to develop systems capable of rendering expressive musical performances from symbolic scores in MusicXML format.

This year, we limit the task to be '''solo-piano''' specific. We accept system that generate symbolic (MIDI) or audio (wav) renderings, and the output shall contain human-like expressive deviation from the MusicXML score.

Similar to AI song contest, the evaluation of expressive rendering is subjective and requires human judges to assess the quality of the generated performances. Thus, we propose a two-phase competition structure as shown in the next section, relying on audience voting to determine the winner.

== Competition Structure ==

RenCon 2025 is structured in two phases:

* '''Phase 1 – Preliminary Round (Online)'''
Submit performances of assigned and free-choice pieces. Includes symbolic or audio renderings and a technical report. No real-time constraints allow for broader participation and diversity in evaluation.
The submission period is open from '''May 30, 2025''' to '''Aug 20, 2025''', following the MIREX submission guidelines.
After the submission deadline, the preliminary round page will be finalized with the list of participants and their submissions, and the online evaluation will take place.

* '''Phase 2 – Live Contest at ISMIR (Daejeon, Korea)'''
Top systems from preliminary round will be invited to render a surprise piece live at ISMIR, using their system in real time.
The live contest is open to all ISMIR attendees, as well as general public if the venue allows. The audience will be able to listen to the live performances and vote for their favorite system. The winner will be announced at the end of the conference.

== Timeline ==

* '''May 30, 2025''': Preliminary round submission opens, online judge sign-up opens
* '''Aug 25, 2025''': Preliminary round submission closes (extended dealine)
* '''Aug 25, 2025''': Preliminary round page finalized and online evaluation begins
* '''Sept 10, 2025''': Online evaluation ends, results announced, and top systems invited to live contest
* '''Sept 25, 2025''': Live contest at ISMIR 2025 in Daejeon, Korea

== Submission Requirements ==

The following items are required for submission:

* Code and checkpoint of the system.
* Symbolic (MIDI) or audio (wav) renderings of designated pieces:
** Required pieces (choose 2 out of 4), click to download the MusicXML files:
*** [https://ren-con2025.vercel.app/static/pieces/CAPRICCIO_en_sol_mineur_HWV_483_-_Handel.mxl Handel: Capriccio in G minor, HWV 483]
*** [https://ren-con2025.vercel.app/static/pieces/32_Variations_in_C_minor_WoO_80_First_5.mxl Beethoven: 32 Variations in C minor, WoO 80 - Theme and the first 5 variations]
*** [https://ren-con2025.vercel.app/static/pieces/12_Romances_Op.21__Sergei_Rachmaninoff_Zdes_khorosho_-_Arrangement_for_solo_piano.mxl Rachmaninoff: Здесь хорошо (How Fair this Spot), Op. 21, No. 7 (transcribed for solo piano)]
*** [https://ren-con2025.vercel.app/static/pieces/With_Dog-teams.mxl Amy Beach: Eskimos, Op.64, No.4 - With Dog-Teams]
** One free-choice piece (rendering must be less than 1 minute 30 seconds)
** A total of 3 pieces, with a maximum of 5 minutes of rendering in total.

* Technical report: Please use the template available at [https://ren-con2025.vercel.app/static/RenCon%20Submission%20Report%20Template.zip Submission Report Template (ZIP)]

Final submission must be made through the [[2025_Submission_Guidelines|MIREX submission system]], with maximum zip file of size 5GB.

== Data Format of submission ==

'''Symbolic submissions''': MIDI format with all sound events in program number 1 (solo piano). Track and channel are unrestricted.

'''Audio Submissions''': wav format, 44.1kHz, 16-bit PCM

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we ask for full transparency.

Some suggested datasets for training and validation include:
* '''[https://github.com/fosfrancesco/asap-dataset ASAP]''': A large dataset of classical piano performances sourced from MAESTRO, includes corresponding MIDI and audio. [https://github.com/CPJKU/asap-dataset (n)ASAP] provides score-performance alignment.
* '''[https://github.com/tangjjbetsy/ATEPP ATEPP]''': A large dataset of transcribed MIDI expressive piano performances, organized by virtuosic performer. However, only around half of the dataset contains score MusicXML files.
* '''[https://github.com/CPJKU/vienna4x22 VIENNA 4x22]''': A small-scale dataset of 4 pieces with 22 different interpretations, including audio and MIDI and fine alignment. ()
* '''[https://github.com/huispaty/batik_plays_mozart Batik-plays-Mozart]''': Fine-aligned performance MIDI dataset of Mozart played by Roland Batik.

Please clearly describe the datasets used for training and validation in your technical report. Important details to include are:

* Dataset name or source
* Size and number of pieces
* Instrumentation and expressive characteristics
* Data format (MIDI, audio, etc.)
* Any preprocessing, cleaning, or augmentation steps applied

This helps the jury and the research community understand the representational capacity and limitations of each submission.

== Post-Processing ==

To ensure fair evaluation, all post-processing applied to the preliminary round output must be documented in the submission report. Depending on your system type, please include the following:

* '''Symbolic Output System''': If your model generates symbolic MIDI output and you submit the sonified audio track, describe how audio is derived. Include soundfont names, software synths used (e.g., FluidSynth, Logic Pro), or player piano models.
** If you would like to submit the MIDI output directly and allow us (the organizer team) for sonification, please contact [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk] during your submission. It's likely that we would arrange a Disklavier recording in the Vienna office of Institute of Computational Perception (CPJKU) lab.
* '''Audio Output Systems''': If your model outputs audio directly, describe if you have applied any enhancement steps such as EQ, reverb, compression, or noise reduction to the model's output.
* '''Controllability or Interventions''': Clarify if the output is influenced by human-involved choices — such as selected tempo, dynamics range, segmentation, or annotated phrasing.
* '''MIDI Cleanup''': If symbolic outputs were manually edited (quantization, pedals, etc) before submission, that should be documented.

Submissions should aim for minimal human intervention. Manual correction is allowed only if it is well-documented and justified in the report.

== Organizers ==
* Huan Zhang (Task Captain, Queen Mary University of London)
* Taegyun Kwon, Venue Coordinator, Korea Advanced Institute of Science and Technology (KAIST)
* Junyan Jiang (New York University)
* Simon Dixon (Queen Mary University of London)
* Gus Xia (MBZUAI)
* Akira Maezawa (Yamaha)

Contact: [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]

2025:RenCon

2025-08-19T22:18:23Z

Huanz: /* Timeline */

= RenCon 2025: Expressive Performance Rendering Competition =

Welcome to the official MIREX wiki page for RenCon 2025, a new task on expressive musical rendering from symbolic scores. This task is part of the MIREX 2025 evaluation campaign and will culminate with a live contest at ISMIR 2025 in Daejeon, Korea. For updates, templates, and examples, please visit our official site: https://ren-con2025.vercel.app

Here is a short summary of the important information:

== Overview ==

Welcome to the official site for '''RenCon 2025''', an international challenge where researchers and developers submit systems capable of rendering expressive musical performances from symbolic scores. This year, we are delighted to hosted it along with '''ISMIR 2025''' and under the '''MIREX Tasks protocol'''.

The RenCon competition has a rich history, having record for compeition in 2002, 2003, 2004, 2005, 2008 and 2011 ([https://www.researchgate.net/publication/228715822_The_Second_Rencon_Performance_Contest_Panel_Discussion_and_the_Future 1], [https://www.nime.org/proceedings/2004/nime2004_120.pdf 2], [https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=e611d7c4f9df63d99d5760ca88a7107dca945e05 3], [https://www.researchgate.net/publication/228715816_Rencon_Performance_Rendering_Contest_for_Automated_Music_Systems 4], [http://smc.afim-asso.org/smc11/papers/smc2011_120.pdf 5]), jointly with conferences like SMC, NIME, ICMPC and IJCAI. Even before the term "AI" was widely used, the RenCon competition has been a platform for researchers to showcase their work in the field of expressive performance rendering.
However, not too much of its past can be traced today, except an old [http://renconmusic.org site]. We hope to revive this tradition with RenCon 2025, coinciding with the renewed global focus on performance during this year's Chopin Piano Competition.

== Task Description ==

Expressive Performance Rendering is a task that challenges participants to develop systems capable of rendering expressive musical performances from symbolic scores in MusicXML format.

This year, we limit the task to be '''solo-piano''' specific. We accept system that generate symbolic (MIDI) or audio (wav) renderings, and the output shall contain human-like expressive deviation from the MusicXML score.

Similar to AI song contest, the evaluation of expressive rendering is subjective and requires human judges to assess the quality of the generated performances. Thus, we propose a two-phase competition structure as shown in the next section, relying on audience voting to determine the winner.

== Competition Structure ==

RenCon 2025 is structured in two phases:

* '''Phase 1 – Preliminary Round (Online)'''
Submit performances of assigned and free-choice pieces. Includes symbolic or audio renderings and a technical report. No real-time constraints allow for broader participation and diversity in evaluation.
The submission period is open from '''May 30, 2025''' to '''Aug 20, 2025''', following the MIREX submission guidelines.
After the submission deadline, the preliminary round page will be finalized with the list of participants and their submissions, and the online evaluation will take place.

* '''Phase 2 – Live Contest at ISMIR (Daejeon, Korea)'''
Top systems from preliminary round will be invited to render a surprise piece live at ISMIR, using their system in real time.
The live contest is open to all ISMIR attendees, as well as general public if the venue allows. The audience will be able to listen to the live performances and vote for their favorite system. The winner will be announced at the end of the conference.

== Timeline ==

* '''May 30, 2025''': Preliminary round submission opens, online judge sign-up opens
* '''Aug 20, 2025''': Preliminary round submission closes
* '''Aug 25, 2025''': Preliminary round page finalized and online evaluation begins
* '''Sept 10, 2025''': Online evaluation ends, results announced, and top systems invited to live contest
* '''Sept 25, 2025''': Live contest at ISMIR 2025 in Daejeon, Korea

== Submission Requirements ==

The following items are required for submission:

* Code and checkpoint of the system.
* Symbolic (MIDI) or audio (wav) renderings of designated pieces:
** Required pieces (choose 2 out of 4), click to download the MusicXML files:
*** [https://ren-con2025.vercel.app/static/pieces/CAPRICCIO_en_sol_mineur_HWV_483_-_Handel.mxl Handel: Capriccio in G minor, HWV 483]
*** [https://ren-con2025.vercel.app/static/pieces/32_Variations_in_C_minor_WoO_80_First_5.mxl Beethoven: 32 Variations in C minor, WoO 80 - Theme and the first 5 variations]
*** [https://ren-con2025.vercel.app/static/pieces/12_Romances_Op.21__Sergei_Rachmaninoff_Zdes_khorosho_-_Arrangement_for_solo_piano.mxl Rachmaninoff: Здесь хорошо (How Fair this Spot), Op. 21, No. 7 (transcribed for solo piano)]
*** [https://ren-con2025.vercel.app/static/pieces/With_Dog-teams.mxl Amy Beach: Eskimos, Op.64, No.4 - With Dog-Teams]
** One free-choice piece (rendering must be less than 1 minute 30 seconds)
** A total of 3 pieces, with a maximum of 5 minutes of rendering in total.

* Technical report: Please use the template available at [https://ren-con2025.vercel.app/static/RenCon%20Submission%20Report%20Template.zip Submission Report Template (ZIP)]

Final submission must be made through the [[2025_Submission_Guidelines|MIREX submission system]], with maximum zip file of size 5GB.

== Data Format of submission ==

'''Symbolic submissions''': MIDI format with all sound events in program number 1 (solo piano). Track and channel are unrestricted.

'''Audio Submissions''': wav format, 44.1kHz, 16-bit PCM

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we ask for full transparency.

Some suggested datasets for training and validation include:
* '''[https://github.com/fosfrancesco/asap-dataset ASAP]''': A large dataset of classical piano performances sourced from MAESTRO, includes corresponding MIDI and audio. [https://github.com/CPJKU/asap-dataset (n)ASAP] provides score-performance alignment.
* '''[https://github.com/tangjjbetsy/ATEPP ATEPP]''': A large dataset of transcribed MIDI expressive piano performances, organized by virtuosic performer. However, only around half of the dataset contains score MusicXML files.
* '''[https://github.com/CPJKU/vienna4x22 VIENNA 4x22]''': A small-scale dataset of 4 pieces with 22 different interpretations, including audio and MIDI and fine alignment. ()
* '''[https://github.com/huispaty/batik_plays_mozart Batik-plays-Mozart]''': Fine-aligned performance MIDI dataset of Mozart played by Roland Batik.

Please clearly describe the datasets used for training and validation in your technical report. Important details to include are:

* Dataset name or source
* Size and number of pieces
* Instrumentation and expressive characteristics
* Data format (MIDI, audio, etc.)
* Any preprocessing, cleaning, or augmentation steps applied

This helps the jury and the research community understand the representational capacity and limitations of each submission.

== Post-Processing ==

To ensure fair evaluation, all post-processing applied to the preliminary round output must be documented in the submission report. Depending on your system type, please include the following:

* '''Symbolic Output System''': If your model generates symbolic MIDI output and you submit the sonified audio track, describe how audio is derived. Include soundfont names, software synths used (e.g., FluidSynth, Logic Pro), or player piano models.
** If you would like to submit the MIDI output directly and allow us (the organizer team) for sonification, please contact [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk] during your submission. It's likely that we would arrange a Disklavier recording in the Vienna office of Institute of Computational Perception (CPJKU) lab.
* '''Audio Output Systems''': If your model outputs audio directly, describe if you have applied any enhancement steps such as EQ, reverb, compression, or noise reduction to the model's output.
* '''Controllability or Interventions''': Clarify if the output is influenced by human-involved choices — such as selected tempo, dynamics range, segmentation, or annotated phrasing.
* '''MIDI Cleanup''': If symbolic outputs were manually edited (quantization, pedals, etc) before submission, that should be documented.

Submissions should aim for minimal human intervention. Manual correction is allowed only if it is well-documented and justified in the report.

== Organizers ==
* Huan Zhang (Task Captain, Queen Mary University of London)
* Taegyun Kwon, Venue Coordinator, Korea Advanced Institute of Science and Technology (KAIST)
* Junyan Jiang (New York University)
* Simon Dixon (Queen Mary University of London)
* Gus Xia (MBZUAI)
* Akira Maezawa (Yamaha)

Contact: [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]

2025:RenCon

2025-08-19T22:17:45Z

Huanz: /* Organizers */

= RenCon 2025: Expressive Performance Rendering Competition =

Welcome to the official MIREX wiki page for RenCon 2025, a new task on expressive musical rendering from symbolic scores. This task is part of the MIREX 2025 evaluation campaign and will culminate with a live contest at ISMIR 2025 in Daejeon, Korea. For updates, templates, and examples, please visit our official site: https://ren-con2025.vercel.app

Here is a short summary of the important information:

== Overview ==

Welcome to the official site for '''RenCon 2025''', an international challenge where researchers and developers submit systems capable of rendering expressive musical performances from symbolic scores. This year, we are delighted to hosted it along with '''ISMIR 2025''' and under the '''MIREX Tasks protocol'''.

The RenCon competition has a rich history, having record for compeition in 2002, 2003, 2004, 2005, 2008 and 2011 ([https://www.researchgate.net/publication/228715822_The_Second_Rencon_Performance_Contest_Panel_Discussion_and_the_Future 1], [https://www.nime.org/proceedings/2004/nime2004_120.pdf 2], [https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=e611d7c4f9df63d99d5760ca88a7107dca945e05 3], [https://www.researchgate.net/publication/228715816_Rencon_Performance_Rendering_Contest_for_Automated_Music_Systems 4], [http://smc.afim-asso.org/smc11/papers/smc2011_120.pdf 5]), jointly with conferences like SMC, NIME, ICMPC and IJCAI. Even before the term "AI" was widely used, the RenCon competition has been a platform for researchers to showcase their work in the field of expressive performance rendering.
However, not too much of its past can be traced today, except an old [http://renconmusic.org site]. We hope to revive this tradition with RenCon 2025, coinciding with the renewed global focus on performance during this year's Chopin Piano Competition.

== Task Description ==

Expressive Performance Rendering is a task that challenges participants to develop systems capable of rendering expressive musical performances from symbolic scores in MusicXML format.

This year, we limit the task to be '''solo-piano''' specific. We accept system that generate symbolic (MIDI) or audio (wav) renderings, and the output shall contain human-like expressive deviation from the MusicXML score.

Similar to AI song contest, the evaluation of expressive rendering is subjective and requires human judges to assess the quality of the generated performances. Thus, we propose a two-phase competition structure as shown in the next section, relying on audience voting to determine the winner.

== Competition Structure ==

RenCon 2025 is structured in two phases:

* '''Phase 1 – Preliminary Round (Online)'''
Submit performances of assigned and free-choice pieces. Includes symbolic or audio renderings and a technical report. No real-time constraints allow for broader participation and diversity in evaluation.
The submission period is open from '''May 30, 2025''' to '''Aug 20, 2025''', following the MIREX submission guidelines.
After the submission deadline, the preliminary round page will be finalized with the list of participants and their submissions, and the online evaluation will take place.

* '''Phase 2 – Live Contest at ISMIR (Daejeon, Korea)'''
Top systems from preliminary round will be invited to render a surprise piece live at ISMIR, using their system in real time.
The live contest is open to all ISMIR attendees, as well as general public if the venue allows. The audience will be able to listen to the live performances and vote for their favorite system. The winner will be announced at the end of the conference.

== Timeline ==

* '''May 30, 2025''': Preliminary round submission opens, online judge sign-up opens
* '''Aug 20, 2025''': Preliminary round submission closes
* '''Aug 25, 2025''': Preliminary round page finalized and online evaluation begins
* '''Sept 10, 2025''': Online evaluation ends, results announced, and top systems invited to live contest
* '''Sept 2*, 2025(TBD)''': Live contest at ISMIR 2025 in Daejeon, Korea

== Submission Requirements ==

The following items are required for submission:

* Code and checkpoint of the system.
* Symbolic (MIDI) or audio (wav) renderings of designated pieces:
** Required pieces (choose 2 out of 4), click to download the MusicXML files:
*** [https://ren-con2025.vercel.app/static/pieces/CAPRICCIO_en_sol_mineur_HWV_483_-_Handel.mxl Handel: Capriccio in G minor, HWV 483]
*** [https://ren-con2025.vercel.app/static/pieces/32_Variations_in_C_minor_WoO_80_First_5.mxl Beethoven: 32 Variations in C minor, WoO 80 - Theme and the first 5 variations]
*** [https://ren-con2025.vercel.app/static/pieces/12_Romances_Op.21__Sergei_Rachmaninoff_Zdes_khorosho_-_Arrangement_for_solo_piano.mxl Rachmaninoff: Здесь хорошо (How Fair this Spot), Op. 21, No. 7 (transcribed for solo piano)]
*** [https://ren-con2025.vercel.app/static/pieces/With_Dog-teams.mxl Amy Beach: Eskimos, Op.64, No.4 - With Dog-Teams]
** One free-choice piece (rendering must be less than 1 minute 30 seconds)
** A total of 3 pieces, with a maximum of 5 minutes of rendering in total.

* Technical report: Please use the template available at [https://ren-con2025.vercel.app/static/RenCon%20Submission%20Report%20Template.zip Submission Report Template (ZIP)]

Final submission must be made through the [[2025_Submission_Guidelines|MIREX submission system]], with maximum zip file of size 5GB.

== Data Format of submission ==

'''Symbolic submissions''': MIDI format with all sound events in program number 1 (solo piano). Track and channel are unrestricted.

'''Audio Submissions''': wav format, 44.1kHz, 16-bit PCM

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we ask for full transparency.

Some suggested datasets for training and validation include:
* '''[https://github.com/fosfrancesco/asap-dataset ASAP]''': A large dataset of classical piano performances sourced from MAESTRO, includes corresponding MIDI and audio. [https://github.com/CPJKU/asap-dataset (n)ASAP] provides score-performance alignment.
* '''[https://github.com/tangjjbetsy/ATEPP ATEPP]''': A large dataset of transcribed MIDI expressive piano performances, organized by virtuosic performer. However, only around half of the dataset contains score MusicXML files.
* '''[https://github.com/CPJKU/vienna4x22 VIENNA 4x22]''': A small-scale dataset of 4 pieces with 22 different interpretations, including audio and MIDI and fine alignment. ()
* '''[https://github.com/huispaty/batik_plays_mozart Batik-plays-Mozart]''': Fine-aligned performance MIDI dataset of Mozart played by Roland Batik.

Please clearly describe the datasets used for training and validation in your technical report. Important details to include are:

* Dataset name or source
* Size and number of pieces
* Instrumentation and expressive characteristics
* Data format (MIDI, audio, etc.)
* Any preprocessing, cleaning, or augmentation steps applied

This helps the jury and the research community understand the representational capacity and limitations of each submission.

== Post-Processing ==

To ensure fair evaluation, all post-processing applied to the preliminary round output must be documented in the submission report. Depending on your system type, please include the following:

* '''Symbolic Output System''': If your model generates symbolic MIDI output and you submit the sonified audio track, describe how audio is derived. Include soundfont names, software synths used (e.g., FluidSynth, Logic Pro), or player piano models.
** If you would like to submit the MIDI output directly and allow us (the organizer team) for sonification, please contact [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk] during your submission. It's likely that we would arrange a Disklavier recording in the Vienna office of Institute of Computational Perception (CPJKU) lab.
* '''Audio Output Systems''': If your model outputs audio directly, describe if you have applied any enhancement steps such as EQ, reverb, compression, or noise reduction to the model's output.
* '''Controllability or Interventions''': Clarify if the output is influenced by human-involved choices — such as selected tempo, dynamics range, segmentation, or annotated phrasing.
* '''MIDI Cleanup''': If symbolic outputs were manually edited (quantization, pedals, etc) before submission, that should be documented.

Submissions should aim for minimal human intervention. Manual correction is allowed only if it is well-documented and justified in the report.

== Organizers ==
* Huan Zhang (Task Captain, Queen Mary University of London)
* Taegyun Kwon, Venue Coordinator, Korea Advanced Institute of Science and Technology (KAIST)
* Junyan Jiang (New York University)
* Simon Dixon (Queen Mary University of London)
* Gus Xia (MBZUAI)
* Akira Maezawa (Yamaha)

Contact: [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]

2025:RenCon

2025-05-21T22:34:33Z

Huanz:

= RenCon 2025: Expressive Performance Rendering Competition =

Welcome to the official MIREX wiki page for RenCon 2025, a new task on expressive musical rendering from symbolic scores. This task is part of the MIREX 2025 evaluation campaign and will culminate with a live contest at ISMIR 2025 in Daejeon, Korea. For updates, templates, and examples, please visit our official site: https://ren-con2025.vercel.app

Here is a short summary of the important information:

== Overview ==

Welcome to the official site for '''RenCon 2025''', an international challenge where researchers and developers submit systems capable of rendering expressive musical performances from symbolic scores. This year, we are delighted to hosted it along with '''ISMIR 2025''' and under the '''MIREX Tasks protocol'''.

The RenCon competition has a rich history, having record for compeition in 2002, 2003, 2004, 2005, 2008 and 2011 ([https://www.researchgate.net/publication/228715822_The_Second_Rencon_Performance_Contest_Panel_Discussion_and_the_Future 1], [https://www.nime.org/proceedings/2004/nime2004_120.pdf 2], [https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=e611d7c4f9df63d99d5760ca88a7107dca945e05 3], [https://www.researchgate.net/publication/228715816_Rencon_Performance_Rendering_Contest_for_Automated_Music_Systems 4], [http://smc.afim-asso.org/smc11/papers/smc2011_120.pdf 5]), jointly with conferences like SMC, NIME, ICMPC and IJCAI. Even before the term "AI" was widely used, the RenCon competition has been a platform for researchers to showcase their work in the field of expressive performance rendering.
However, not too much of its past can be traced today, except an old [http://renconmusic.org site]. We hope to revive this tradition with RenCon 2025, coinciding with the renewed global focus on performance during this year's Chopin Piano Competition.

== Task Description ==

Expressive Performance Rendering is a task that challenges participants to develop systems capable of rendering expressive musical performances from symbolic scores in MusicXML format.

This year, we limit the task to be '''solo-piano''' specific. We accept system that generate symbolic (MIDI) or audio (wav) renderings, and the output shall contain human-like expressive deviation from the MusicXML score.

Similar to AI song contest, the evaluation of expressive rendering is subjective and requires human judges to assess the quality of the generated performances. Thus, we propose a two-phase competition structure as shown in the next section, relying on audience voting to determine the winner.

== Competition Structure ==

RenCon 2025 is structured in two phases:

* '''Phase 1 – Preliminary Round (Online)'''
Submit performances of assigned and free-choice pieces. Includes symbolic or audio renderings and a technical report. No real-time constraints allow for broader participation and diversity in evaluation.
The submission period is open from '''May 30, 2025''' to '''Aug 20, 2025''', following the MIREX submission guidelines.
After the submission deadline, the preliminary round page will be finalized with the list of participants and their submissions, and the online evaluation will take place.

* '''Phase 2 – Live Contest at ISMIR (Daejeon, Korea)'''
Top systems from preliminary round will be invited to render a surprise piece live at ISMIR, using their system in real time.
The live contest is open to all ISMIR attendees, as well as general public if the venue allows. The audience will be able to listen to the live performances and vote for their favorite system. The winner will be announced at the end of the conference.

== Timeline ==

* '''May 30, 2025''': Preliminary round submission opens, online judge sign-up opens
* '''Aug 20, 2025''': Preliminary round submission closes
* '''Aug 25, 2025''': Preliminary round page finalized and online evaluation begins
* '''Sept 10, 2025''': Online evaluation ends, results announced, and top systems invited to live contest
* '''Sept 2*, 2025(TBD)''': Live contest at ISMIR 2025 in Daejeon, Korea

== Submission Requirements ==

The following items are required for submission:

* Code and checkpoint of the system.
* Symbolic (MIDI) or audio (wav) renderings of designated pieces:
** Required pieces (choose 2 out of 4), click to download the MusicXML files:
*** [https://ren-con2025.vercel.app/static/pieces/CAPRICCIO_en_sol_mineur_HWV_483_-_Handel.mxl Handel: Capriccio in G minor, HWV 483]
*** [https://ren-con2025.vercel.app/static/pieces/32_Variations_in_C_minor_WoO_80_First_5.mxl Beethoven: 32 Variations in C minor, WoO 80 - Theme and the first 5 variations]
*** [https://ren-con2025.vercel.app/static/pieces/12_Romances_Op.21__Sergei_Rachmaninoff_Zdes_khorosho_-_Arrangement_for_solo_piano.mxl Rachmaninoff: Здесь хорошо (How Fair this Spot), Op. 21, No. 7 (transcribed for solo piano)]
*** [https://ren-con2025.vercel.app/static/pieces/With_Dog-teams.mxl Amy Beach: Eskimos, Op.64, No.4 - With Dog-Teams]
** One free-choice piece (rendering must be less than 1 minute 30 seconds)
** A total of 3 pieces, with a maximum of 5 minutes of rendering in total.

* Technical report: Please use the template available at [https://ren-con2025.vercel.app/static/RenCon%20Submission%20Report%20Template.zip Submission Report Template (ZIP)]

Final submission must be made through the [[2025_Submission_Guidelines|MIREX submission system]], with maximum zip file of size 5GB.

== Data Format of submission ==

'''Symbolic submissions''': MIDI format with all sound events in program number 1 (solo piano). Track and channel are unrestricted.

'''Audio Submissions''': wav format, 44.1kHz, 16-bit PCM

== Training Datasets ==

Participants are welcome to train their systems on any dataset, including publicly available corpora, proprietary collections, or internally curated material. There are no restrictions on dataset origin, but we ask for full transparency.

Some suggested datasets for training and validation include:
* '''[https://github.com/fosfrancesco/asap-dataset ASAP]''': A large dataset of classical piano performances sourced from MAESTRO, includes corresponding MIDI and audio. [https://github.com/CPJKU/asap-dataset (n)ASAP] provides score-performance alignment.
* '''[https://github.com/tangjjbetsy/ATEPP ATEPP]''': A large dataset of transcribed MIDI expressive piano performances, organized by virtuosic performer. However, only around half of the dataset contains score MusicXML files.
* '''[https://github.com/CPJKU/vienna4x22 VIENNA 4x22]''': A small-scale dataset of 4 pieces with 22 different interpretations, including audio and MIDI and fine alignment. ()
* '''[https://github.com/huispaty/batik_plays_mozart Batik-plays-Mozart]''': Fine-aligned performance MIDI dataset of Mozart played by Roland Batik.

Please clearly describe the datasets used for training and validation in your technical report. Important details to include are:

* Dataset name or source
* Size and number of pieces
* Instrumentation and expressive characteristics
* Data format (MIDI, audio, etc.)
* Any preprocessing, cleaning, or augmentation steps applied

This helps the jury and the research community understand the representational capacity and limitations of each submission.

== Post-Processing ==

To ensure fair evaluation, all post-processing applied to the preliminary round output must be documented in the submission report. Depending on your system type, please include the following:

* '''Symbolic Output System''': If your model generates symbolic MIDI output and you submit the sonified audio track, describe how audio is derived. Include soundfont names, software synths used (e.g., FluidSynth, Logic Pro), or player piano models.
** If you would like to submit the MIDI output directly and allow us (the organizer team) for sonification, please contact [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk] during your submission. It's likely that we would arrange a Disklavier recording in the Vienna office of Institute of Computational Perception (CPJKU) lab.
* '''Audio Output Systems''': If your model outputs audio directly, describe if you have applied any enhancement steps such as EQ, reverb, compression, or noise reduction to the model's output.
* '''Controllability or Interventions''': Clarify if the output is influenced by human-involved choices — such as selected tempo, dynamics range, segmentation, or annotated phrasing.
* '''MIDI Cleanup''': If symbolic outputs were manually edited (quantization, pedals, etc) before submission, that should be documented.

Submissions should aim for minimal human intervention. Manual correction is allowed only if it is well-documented and justified in the report.

== Organizers ==
* Huan Zhang (Task Captain, Queen Mary University of London)
* Junyan Jiang (New York University )
* Simon Dixon (Queen Mary University of London)
* Gus Xia (MBZUAI)
* Akira Maezawa (Yamaha)

Contact: [mailto:huan.zhang@qmul.ac.uk huan.zhang@qmul.ac.uk]

2025:RenCon

2025-05-18T20:47:43Z

Huanz: overview