# YuanGongND/ast

Code for the Interspeech 2021 paper "AST: Audio Spectrogram Transformer".

Repository: https://github.com/YuanGongND/ast
Canonical: https://ross.abutalabs.com/products/ast
Language: Jupyter Notebook
License: BSD-3-Clause
License Family: permissive
Topics: pytorch, audio-classification, deep-learning, audio, representation-learning, keyword-spotting, speech-commands, speech-classification
Last push: 2023-05-21T21:12:01+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 1918, "days_push": 1200, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1472, forks 244 (observed 2026-08-28T04:04:49.638498+00:00)

## What it is
Official PyTorch implementation of the Audio Spectrogram Transformer (AST) from the Interspeech 2021 paper, which applies a Vision Transformer to audio spectrograms for audio classification. Includes pretrained models and recipes for AudioSet, ESC-50, and Speech Commands datasets.

## Use cases
- classify audio clips with a transformer model
- fine-tune a pretrained audio classification model on my own dataset
- run keyword spotting on speech commands
- evaluate on ESC-50 environmental sound classification
- extract audio representations for downstream tasks
- visualize transformer attention over audio spectrograms

## When to choose
- you need a strong, well-cited transformer baseline for audio classification
- you want pretrained AudioSet models for transfer learning
- you need reproducible recipes for ESC-50, Speech Commands, or AudioSet

## When to avoid
- you need real-time or low-latency audio inference on edge devices
- you want speech recognition or transcription rather than classification
- you need actively maintained code with frequent updates

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, audio-processing, deep-learning
- domain: machine-learning, deep-learning, speech-processing
- platform: python
- tags: audio-classification, transformer, pytorch, audio-spectrogram, keyword-spotting, pretrained-models, research-code, audio, gpu

## Member repositories
- YuanGongND/ast (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:49.638498+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:34:40.245243+00:00, confidence not recorded.
  - readme: https://github.com/YuanGongND/ast (fetched 2026-08-28T04:04:49.638498+00:00, sha a3bde2b8381f)
- Data as of 2026-08-30T08:39:29.467469+00:00.