Wiki Index

Topics

• topics/audio-classification-and-representation — Audio tagging, representation learning, audio-language models
• topics/classical-techniques — Pre-deep-learning foundations: ASA, CASA, NMF, sinusoidal modeling
• topics/datasets — Training/evaluation datasets for separation, transcription, classification
• topics/source-separation — Deep and classical approaches to isolating sources from mixtures
• topics/tools-and-libraries — Software libraries, CLI tools, impulse response collections
• topics/transcription-and-pitch-estimation — AMT, f0 estimation, guitar tab transcription

Concepts

• concepts/auditory-scene-analysis — Bregman's psychological framework for parsing auditory scenes
• concepts/common-fate-principle — Gestalt principle: synchronous changes = one source
• concepts/computational-auditory-scene-analysis — Computational implementation of ASA principles
• concepts/deep-clustering-separation — Embedding-based separation via clustering (Hershey et al., 2016)
• concepts/informed-model-based-separation — Separation guided by score/MIDI/annotations
• concepts/mixture-invariant-training — MixIT: training without isolated stems (Wisdom, Hershey et al.)
• concepts/musicxml-tab-notation — MusicXML representation for guitar tablature
• concepts/non-negative-matrix-factorization-audio — NMF applied to spectrograms (Smaragdis & Brown, 2003)
• concepts/permutation-invariant-training — PIT: handling source permutation ambiguity in training
• concepts/query-based-source-separation — Text-queried / conditional source separation
• concepts/self-supervised-audio-representation — Learning audio embeddings without labels
• concepts/shift-invariant-nmf — Convolutive NMF, sparse NMF, and other NMF extensions
• concepts/sinusoidal-modeling — McAulay-Quatieri sinusoidal analysis/synthesis (1986)
• concepts/spectrogram-unets — U-Net architectures applied to time-frequency representations
• concepts/synthetic-mixing-pipelines — Data augmentation via synthetic mixture creation

Entities

• entities/audiosef — AudioSep: text-conditional source separation
• entities/basic-pitch — Spotify's lightweight audio-to-MIDI (pip installable)
• entities/bs-roformer — Current SOTA source separation (ByteDance, 2023)
• entities/crepe — Monophonic F0 estimation with deep CNN
• entities/demucs — Meta's source separation (Demucs / HDemucs / HTDemucs)
• entities/guitarset — Dataset: solo guitar with per-string annotations
• entities/mt3 — Google's multi-instrument T5-based transcription
• entities/onsets-and-frames — Magenta piano transcription (CNN+BiLSTM)
• entities/open-unmix — Hackable reference separation implementation
• entities/spleeter — Deezer's fast source separation library
• entities/tabcnn — Guitar tablature transcription from audio

Sources (by ingest date, newest first)

• sources/2024-04-19-synthtab — SynthTab: synthetic data for guitar tab (2024) — synthetic data approach to tab transcription
• sources/2023-09-05-bs-roformer — BS-RoFormer (2023) — current SOTA source separation
• sources/2023-08-10-audiosef — AudioSep (2023) — text-conditional separation
• sources/2022-11-15-htdemucs — HTDemucs (2022) — hybrid transformer separation
• sources/2022-03-23-basic-pitch — Basic Pitch (2022) — lightweight audio-to-MIDI
• sources/2019-11-27-demucs — Demucs (2019) — waveform-domain separation
• sources/2019-11-04-tabcnn — TabCNN (2019) — guitar tablature with CNN
• sources/2018-09-23-guitarset — GuitarSet (2018) — guitar transcription dataset
• sources/2018-02-16-crepe — CREPE (2018) — deep pitch estimation
• sources/2017-10-30-onsets-and-frames — Onsets and Frames (2017) — dual-objective piano transcription

Comparisons

No comparison pages yet.