Update README.md

README.md

@@ -1,43 +1,143 @@
 ---
 license: mit
+tags:
+  - symbolic-music
+  - music-information-retrieval
+  - classification
+  - retrieval
+  - benchmark
 ---
 
-**Overview.**
+# SyMuRBench Datasets and Precomputed Features
 
-Here you can find datasets for SyMuRBench and precomputed music21 and jSymbolic features for all tasks.
+This repository contains datasets and precomputed features for [SyMuRBench](https://github.com/Mintas/SyMuRBench), a benchmark for symbolic music understanding models. It includes metadata and MIDI files for multiple classification and retrieval tasks, along with pre-extracted **music21** and **jSymbolic** features.
 
-**Tasks description.**
+You can install and use the full pipeline via:
+👉 [https://github.com/Mintas/SyMuRBench](https://github.com/Mintas/SyMuRBench)
 
-| Task name | Source dataset | Task type | # of classes | # of files | Default metrics |
-| -------- | ------- | ---------- | -------------- | ------- | ------- |
-| ComposerClassificationASAP | ASAP | Multiclass classification | 7 | 197 | weighted f1 score, balanced accuracy |
-| GenreClassificationMMD | MetaMIDI | Multiclass classification | 7 | 2795 | weighted f1 score, balanced accuracy |
-| GenreClassificationWMTX | WikiMT-X | Multiclass classification | 8 | 985 | weighted f1 score, balanced accuracy |
-| EmotionClassificationEMOPIA | Emopia | Multiclass classification | 4 | 191 | weighted f1 score, balanced accuracy |
-| EmotionClassificaitonMIREX | MIREX | Multiclass classification | 5 | 163 | weighted f1 score, balanced accuracy |
-| InstrumentDetectionMMD | MetaMIDI | Multilabel classification | 128 | 4675 | weighted f1 score |
-| ScorePerfomanceRetrievalASAP | ASAP | Retrieval | - | 438 (219 pairs) | R@1, R@5, R@10, Median Rank |
+---
+
+## Overview
+
+SyMuRBench supports evaluation across diverse symbolic music tasks, including composer, genre, emotion, and instrument classification, as well as score-performance retrieval. This Hugging Face dataset provides:
+
+- Dataset metadata (CSV files)
+- MIDI files organized by task
+- Precomputed **music21** and **jSymbolic** features
+- Configuration-ready structure for immediate use in benchmarking
+
+---
+
+## Tasks Description
 
+| Task Name | Source Dataset | Task Type | # of Classes | # of Files | Default Metrics |
+|----------|----------------|-----------|--------------|------------|-----------------|
+| ComposerClassificationASAP | ASAP | Multiclass Classification | 7 | 197 | weighted f1 score, balanced accuracy |
+| GenreClassificationMMD | MetaMIDI | Multiclass Classification | 7 | 2,795 | weighted f1 score, balanced accuracy |
+| GenreClassificationWMTX | WikiMT-X | Multiclass Classification | 8 | 985 | weighted f1 score, balanced accuracy |
+| EmotionClassificationEMOPIA | Emopia | Multiclass Classification | 4 | 191 | weighted f1 score, balanced accuracy |
+| EmotionClassificationMIREX | MIREX | Multiclass Classification | 5 | 163 | weighted f1 score, balanced accuracy |
+| InstrumentDetectionMMD | MetaMIDI | Multilabel Classification | 128 | 4,675 | weighted f1 score |
+| ScorePerformanceRetrievalASAP | ASAP | Retrieval | - | 438 (219 pairs) | R@1, R@5, R@10, Median Rank |
 
-**Precomputed features.**
+---
+
+## Precomputed Features
 
-The pre-calculated features can be found in the "data/features" folder within the files named "music21_full_dataset.parquet" and "jsymbolic_full_dataset.parquet".
+Precomputed features are available in the `data/features/` folder:
+- `music21_full_dataset.parquet`
+- `jsymbolic_full_dataset.parquet`
 
-Below is an example of a table containing these features:
+Each file contains a unified table with:
+- `midi_file`: Filename of the MIDI
+- `task`: Corresponding task name
+- `E_0` to `E_N`: Feature vector
 
-| midi_file | task | E_0 | E_1 | ... | E_631 | E_632 |
-| --- | --- | --- | --- | --- | --- | --- |
+### Example
+
+| midi_file | task | E_0 | E_1 | ... | E_672 | E_673 |
+|----------|------|-----|-----|-----|-------|-------|
 | Q1_0vLPYiPN7qY_1.mid | EmotionClassificationEMOPIA | 0.0 | 0.0 | ... | 0.0 | 0.0 |
-| Q1_0vLPYiPN7qY_2.mid | EmotionClassificationEMOPIA | 0.0 | 0.0 | ... | 0.0 | 0.0 |
 | Q1_4dXC1cC7crw_0.mid | EmotionClassificationEMOPIA | 0.0 | 0.0 | ... | 0.0 | 0.0 |
-| Q1_4dXC1cC7crw_1.mid | EmotionClassificationEMOPIA | 0.0 | 0.0 | ... | 0.0 | 0.0 |
-| Q1_GY3f6ckBVkA_2.mid | EmotionClassificationEMOPIA | 0.0 | 0.0 | ... | 0.0 | 0.0 |
 
-**File structure.**
+These can be loaded with:
+```python
+import pandas as pd
+df = pd.read_parquet("data/features/music21_full_dataset.parquet")
+```
+
+## File Structure
+
+The dataset is distributed as a ZIP archive:
+
+`data/datasets.zip`
+
+After extraction, the structure is:
 
-The datasets file structure consists of a "data/datasets.zip" archive containing folder with several subfolders, including "composer_and_retrieval_datasets," "genre_dataset," "wikimtx_dataset," "mirex_dataset," "emopia_dataset," and "instrument_dataset."
+```
+datasets/
+├── composer_and_retrieval_datasets/
+│   ├── metadata_composer_dataset.csv
+│   ├── metadata_retrieval_dataset.csv
+��   └── ... (MIDI files organized in subfolders)
+├── genre_dataset/
+│   ├── metadata_genre_dataset.csv
+│   └── midis/
+├── wikimtx_dataset/
+│   ├── metadata_wikimtx_dataset.csv
+│   └── midis/
+├── emopia_dataset/
+│   ├── metadata_emopia_dataset.csv
+│   └── midis/
+├── mirex_dataset/
+│   ├── metadata_mirex_dataset.csv
+│   └── midis/
+└── instrument_dataset/
+    ├── metadata_instrument_dataset.csv
+    └── midis/
+```
+* CSV files: Contain `filename` and `label` (or pair info for retrieval).
+* MIDI files: Used as input for feature extractors.
+
+---
+
+## How to Use
+
+You can download and extract everything using the built-in utility:
+
+```python
+from symurbench.utils import load_datasets
+
+load_datasets(output_folder="./data", load_features=True)
+```
+
+This will:
+
+* Download datasets.zip and extract it
+* Optionally download precomputed features
+* Update config paths automatically
+
+---
+
+## License
+
+This dataset is released under the MIT License.
+
+---
 
-The "composer_and_retrieval_datasets" folder contains two .csv files with metadata for the tasks "ComposerClassificationASAP" and "ScorePerformanceRetrievalASAP," as well as 16 folders with .mid files organized in a tree-like structure.
+## Citation
 
-The remaining folders each contain a .csv file with metadata and a "midis" folder with .mid files for the corresponding task. For classification tasks, the metadata includes file names and their respective labels. For the "ScorePerformanceRetrievalASAP" task, the metadata includes file names for score-performance pairs and information about duration changes made in the MIDI scores.
+If you use SyMuRBench in your work, please cite:
 
+```bibtex
+@inproceedings{symurbench2025,
+  author    = {Petr Strepetov and Dmitrii Kovalev},
+  title     = {SyMuRBench: Benchmark for Symbolic Music Representations},
+  booktitle = {Proceedings of the 3rd International Workshop on Multimedia Content Generation and Evaluation: New Methods and Practice (McGE '25)},
+  year      = {2025},
+  pages     = {9},
+  publisher = {ACM},
+  address   = {Dublin, Ireland},
+  doi       = {10.1145/3746278.3759392}
+}
+```