# clovaai/voxceleb_trainer

In defence of metric learning for speaker recognition

Repository: https://github.com/clovaai/voxceleb_trainer
Canonical: https://ross.abutalabs.com/products/voxceleb_trainer
Language: Python
License: MIT
License Family: permissive
Topics: speaker-recognition, metric-learning, voxceleb, speaker-verification
Last push: 2026-04-22T05:16:43+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 78, release rhythm 35, longevity 100
- inputs: {"age_days": 2351, "days_push": 133, "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 1175, forks 290 (observed 2026-08-28T04:03:52.473076+00:00)

## What it is
A PyTorch framework for training and evaluating speaker recognition and verification models on the VoxCeleb datasets. It implements multiple network architectures (ResNet, RawNet3, VGGVox) and metric learning loss functions (AM-Softmax, AAM-Softmax, Angular Prototypical, GE2E, Triplet), with pretrained models provided.

## Use cases
- train a speaker verification model on VoxCeleb
- evaluate speaker recognition models with EER metrics
- compare metric learning loss functions for speaker embeddings
- download and run pretrained speaker recognition models
- prepare and augment VoxCeleb audio data for training
- extract speaker embeddings from raw waveforms

## When to choose
- you need to train or benchmark speaker verification models on VoxCeleb
- you want to experiment with metric learning losses like AAM-Softmax or Angular Prototypical
- you need reproducible baselines with pretrained models and published EER scores

## When to avoid
- you need production-ready speaker recognition as a service rather than a research training framework
- your task is general speech recognition or transcription rather than speaker identity
- you work outside Python/PyTorch environments

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, audio-processing, speech-recognition, benchmarking
- domain: machine-learning, deep-learning, speech-processing, artificial-intelligence
- platform: python
- tags: speaker-verification, speaker-recognition, metric-learning, voxceleb, pytorch, research-code, pretrained-models, audio, linux, gpu

## Member repositories
- clovaai/voxceleb_trainer (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:52.473076+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 2026-08-30T06:27:04.470612+00:00, confidence not recorded.
  - readme: https://github.com/clovaai/voxceleb_trainer (fetched 2026-08-28T04:03:52.473076+00:00, sha 2131dfbf70ce)
- Data as of 2026-08-30T08:39:29.467469+00:00.