2024:Symbolic Music Generation - Revision history

Zizzi wang: /* Algorithm Submission */

2025-07-09T07:02:59Z

‎Algorithm Submission

Zhaojw1998: /* References */

2024-10-01T02:10:14Z

‎References

Zhaojw1998: /* Baselines */

2024-10-01T02:07:51Z

‎Baselines

Zhaojw1998: /* Baselines */

2024-10-01T02:06:25Z

‎Baselines

Zhaojw1998: /* References */

2024-09-16T18:22:43Z

‎References

Zhaojw1998: /* Baselines */

2024-09-16T18:07:19Z

‎Baselines

Zhaojw1998: /* Baselines */

2024-09-16T18:05:47Z

‎Baselines

Zhaojw1998: /* Baselines */

2024-09-16T18:04:39Z

‎Baselines

Zizzi wang: /* Description */

2024-09-15T14:21:11Z

‎Description

Zizzi wang: /* References */

2024-09-15T10:18:30Z

‎References

@@ Line 135: / Line 135: @@
 * Shared google drive links
 * If the repository is private, an access token is also required
 =Baselines=

@@ Line 160: / Line 160: @@
 * Lu, P., et al. Musecoco: Generating symbolic music from text. arXiv preprint arXiv:2306.00110 (2023).
 * Wang, Z., et al. Whole-song hierarchical generation of symbolic music using cascaded diffusion models, in ICLR 2024.
 * Zhao, J., & Xia, G. Accomontage: Accompaniment arrangement via phrase selection and style transfer, in ISMIR 2021.
 * Code and data format samples: [https://github.com/ZZWaang/acc-gen-8bar-wholesong/tree/main https://github.com/ZZWaang/acc-gen-8bar-wholesong/tree/main]

@@ Line 141: / Line 141: @@
 To establish a benchmark for this task, we consider the three baseline models in their official implementations:
-'''WholeSongGen''' (Wang et al., 2024), a denoising diffusion probabilistic model (DDPM) generating piano accompaniments as piano-roll images.
+'''WholeSongGen''' (Wang et al., 2024)
-'''Compose & Embellish''' (Wu and Yang, 2023), a Transformer-based architecture generating piano performances in beat-based event sequences.
+'''AccoMontage''' (Zhao and Xia, 2021)
+* A hybrid algorithm generating piano accompaniments by rule-based search and music representation learning.
 =Contacts=

@@ Line 139: / Line 139: @@
 =Baselines=
-To establish a benchmark for this task, we have several options for baseline models, subjective to the availability of official implementations:
+To establish a benchmark for this task, we consider the three baseline models in their official implementations:
-'''WholeSongGen''' (Wang et al., 2024)
+'''WholeSongGen''' (Wang et al., 2024), a denoising diffusion probabilistic model (DDPM) generating piano accompaniments as piano-roll images.
-'''AccoMontage''' (Zhao et al., 2021)
+'''Compose & Embellish''' (Wu and Yang, 2023), a Transformer-based architecture generating piano performances in beat-based event sequences.
-'''Compound Word Transformer''' (Hsiao et al., 2021)
+'''AccoMontage''' (Zhao and Xia, 2021), a hybrid algorithm generating piano accompaniments by rule-based search and music representation learning.
 =Contacts=

@@ Line 168: / Line 168: @@
 =References=
-* Harte, C. (2010). Towards automatic extraction of harmony information from music signals (Doctoral dissertation).* Lu, P., Xu, X., Kang, C., Yu, B., Xing, C., Tan, X., & Bian, J. (2023). Musecoco: Generating symbolic music from text. arXiv preprint arXiv:2306.00110.
+* Harte, C. Towards automatic extraction of harmony information from music signals. PhD Diss. 2010.
-* Wang, Z., Min, L., & Xia, G. Whole-Song Hierarchical Generation of Symbolic Music Using Cascaded Diffusion Models. In The Twelfth International Conference on Learning Representations.
+* Lu, P., et al. Musecoco: Generating symbolic music from text. arXiv preprint arXiv:2306.00110 (2023).
-* Jingwei Zhao, & Gus Xia (2021). AccoMontage: Accompaniment Arrangement via Phrase Selection and Style Transfer. In Proceedings of the 22nd International Society for Music Information Retrieval Conference, ISMIR 2021, Online, November 7-12, 2021 (pp. 833–840).
+* Wang, Z., et al. Whole-song hierarchical generation of symbolic music using cascaded diffusion models, in ICLR 2024.
-* Thickstun, J., Hall, D., Donahue, C., & Liang, P. (2023). Anticipatory music transformer. arXiv preprint arXiv:2306.08620.
+* Zhao, J., & Xia, G. Accomontage: Accompaniment arrangement via phrase selection and style transfer, in ISMIR 2021.
 * Code and data format samples: [https://github.com/ZZWaang/acc-gen-8bar-wholesong/tree/main https://github.com/ZZWaang/acc-gen-8bar-wholesong/tree/main]

@@ Line 145: / Line 145: @@
 This model generates piano accompaniment using a diffusion model.
-'''AccoMontage''' (Zhao et al., 2020)
+'''AccoMontage''' (Zhao et al., 2021)
 This algorithm generates piano accompaniment using a combination of rule-based search and deep representation learning.

@@ Line 139: / Line 139: @@
 =Baselines=
-'''WholeSongGen''' (Wang et al., 2023)
+To establish a benchmark for this task, we have several baseline options, subjective to the availability of official implementations:
-This model generates accompaniment using the diffusion model.
+'''WholeSongGen''' (Wang et al., 2024)
-'''Accomontage''' (Zhao et al., 2020)
+This model generates piano accompaniment using a diffusion model.
-This algorithm generates accompaniment using a combination of rule-based search and deep representation learning.
+'''AccoMontage''' (Zhao et al., 2020)
 '''Anticipatory Music Transformer''' (Thickstun et al., 2024)
-This model generates accompaniment using a Transformer-based model.
+A Transformer-based model for track infilling and accompaniment generation. While the current implementation does not explicitly consider chord input, it remains a relevant starting point to study with.
 =Contacts=

@@ Line 2: / Line 2: @@
 Symbolic music generation is a broad topic. It covers a wide range of tasks, including generation, harmonization, arrangement, instrumentation, and more. We have multiple ways to represent music data, and the evaluation metrics also vary. To define a MIREX challenge within this topic, we need to narrow our focus to specific subtasks that are both relevant to the community and feasible to evaluate effectively.
-This year, we select the task to be '''piano accompaniment arrangement from a lead sheet'''. The lead sheet provides information about the melody, chord progression, and optional phrase labels. The goal is to generate a piano accompaniment that complements the lead melody. The music data consists of 8-measure segments in 4/4 meter, quantized to a sixteenth-note resolution. A more detailed description of the data structure is provided in the data format section. The genre of the lead sheets is broadly within western pop music (refer to the music examples for more detail).
+This year, we select the task to be '''piano accompaniment arrangement from a lead sheet'''. The lead sheet provides information about the melody, and chord progression. The goal is to generate a piano accompaniment that complements the lead melody. The music data consists of 8-measure segments in 4/4 meter, quantized to a sixteenth-note resolution. A more detailed description of the data structure is provided in the data format section. The genre of the lead sheets is broadly within western pop music (refer to the music examples for more detail).
 =Data Format=