$ show Music-and-Culture-Technology-Lab/omnizart
Omnizart is an automatic music transcription (AMT) library that converts polyphonic audio into symbolic representations including pitched notes, drums, chords, vocal melody, and beat/downbeat. Built as a Python CLI tool with multiple subcommands (music, drum, chord, vocal, vocal-contour, beat, patch-cnn), it provides pre-trained TensorFlow models and supports custom training and feature extraction. The architecture follows a modular pattern with separate app, inference, features, labels, and prediction modules per task, all inheriting from common base classes (BaseTranscription, BaseDatasetLoader). While primarily an application library for end-to-end transcription, it has strong research components: it uses standard MIR datasets (MAESTRO, MusicNet, MedleyDB, etc.), implements state-of-the-art models (U-Net with ASPP, attention-based U-Net, PyramidNet, Chord Transformer), and includes custom feature extraction (CFP, CQT, chroma). The repository also bundles a JavaScript waveform playlist library for documentation demos. The codebase shows active maintenance (recent commits in 2026, CI testing up to Python 3.14) but has technical drift risks due to heavy reliance on TensorFlow 1.x compat layers, external dependencies (madmom, vamp), and known bugs in drum model training.
The LLM identified 3 files worth highlighting for reuse:
The LazyLoader class defers module imports until first access, reducing startup time and avoiding hard dependency failures. It is a self-contained, generic utility that can be extracted to any Python project needing lazy imports. It demonstrates non-obvious skill by handling attribute access and submodule imports transparently.
The parallel_generator function wraps ProcessPoolExecutor or ThreadPoolExecutor for parallel iteration over a callable. It is parameterized by number of workers, queue depth, and timeout, making it adaptable to various data processing pipelines. It's isolated (no internal project imports beyond standard libs) and solves the common data engineering problem of parallel feature extraction or I/O-bound tasks.
The json_serializable decorator and associated key_path/value_path pattern provide a declarative way to parse nested YAML/JSON configuration into typed Python classes with automatic schema validation and serialization. This pattern is reusable for any project requiring structured configuration loading, and the decorator is self-contained with no heavy dependencies. It demonstrates a clean separation between configuration parsing and business logic.