# clovaai/aasist architecture

> Architecture diagram of the GitHub repository clovaai/aasist, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-19). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/clovaai/aasist
- Repository: https://github.com/clovaai/aasist

## Overview

AASIST is a Python framework for training and evaluating audio anti-spoofing models on ASVspoof data. The main workflow loads experiment configuration and audio trials, selects a model architecture, then trains or runs evaluation, producing countermeasure scores and EER/t-DCF metrics. The README also documents dataset downloading, pretrained-model evaluation, and four model choices. Source excerpts show the main runtime wiring; some training details and configuration-to-weight mappings are documented but not fully sampled.

Main entry points: [`main.py`](https://github.com/clovaai/aasist/blob/main/main.py) (Experiment runner), [`config`](https://github.com/clovaai/aasist/tree/main/config) (Experiment configs), [`download_dataset.py`](https://github.com/clovaai/aasist/blob/main/download_dataset.py) (Dataset download), [`data_utils.py`](https://github.com/clovaai/aasist/blob/main/data_utils.py) (Trial loading)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_workflow["Experiment workflow"]
  node_main["Experiment runner<br/>[main.py]"]
  node_configs["Experiment configs"]
  node_scores["Score output<br/>[main.py]"]
end

subgraph group_data["Data preparation"]
  node_dataset[("ASVspoof audio")]
  node_download["Dataset download"]
  node_audio_loader["Trial loading<br/>[data_utils.py]"]
end

subgraph group_models["Detection models"]
  node_aasist["AASIST models<br/>[AASIST.py]"]
  node_rawnet2["RawNet2 baseline<br/>[RawNet2Spoof.py]"]
  node_rawgat["RawGAT-ST baseline"]
  node_weights["Pretrained weights"]
end

subgraph group_support["Optimization and metrics"]
  node_optimizer["Training utilities<br/>[utils.py]"]
  node_metrics["Evaluation metrics<br/>[evaluation.py]"]
  node_asv_scores["ASV reference scores"]
end

node_user(("Researcher"))

node_user -->|"runs"| node_main
node_user -->|"requests download"| node_download
node_download -->|"downloads archive"| node_dataset
node_user -->|"selects config"| node_configs
node_configs -->|"configures run"| node_main
node_dataset -->|"provides audio"| node_audio_loader
node_main -->|"builds loaders"| node_audio_loader
node_main -->|"loads architecture"| node_aasist
node_main -.->|"loads architecture"| node_rawnet2
node_main -.->|"loads architecture"| node_rawgat
node_main -->|"loads pretrained weights"| node_weights
node_main -->|"uses training utilities"| node_optimizer
node_main -->|"writes scores"| node_scores
node_main -->|"computes metrics"| node_metrics
node_asv_scores -->|"supplies ASV scores"| node_metrics

click node_main "https://github.com/clovaai/aasist/blob/main/main.py"
click node_configs "https://github.com/clovaai/aasist/tree/main/config"
click node_scores "https://github.com/clovaai/aasist/blob/main/main.py"
click node_download "https://github.com/clovaai/aasist/blob/main/download_dataset.py"
click node_audio_loader "https://github.com/clovaai/aasist/blob/main/data_utils.py"
click node_aasist "https://github.com/clovaai/aasist/blob/main/models/AASIST.py"
click node_rawnet2 "https://github.com/clovaai/aasist/blob/main/models/RawNet2Spoof.py"
click node_rawgat "https://github.com/clovaai/aasist/blob/main/models/RawNetGatSpoofST.py"
click node_weights "https://github.com/clovaai/aasist/tree/main/models/weights"
click node_optimizer "https://github.com/clovaai/aasist/blob/main/utils.py"
click node_metrics "https://github.com/clovaai/aasist/blob/main/evaluation.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_main,node_configs,node_scores toneBlue
class node_dataset,node_download,node_audio_loader toneAmber
class node_aasist,node_rawnet2,node_rawgat,node_weights toneMint
class node_optimizer,node_metrics,node_asv_scores toneRose
class node_user toneIndigo
```

## Components

### Experiment workflow

- [Experiment runner](https://github.com/clovaai/aasist/blob/main/main.py): `main.py`
- [Experiment configs](https://github.com/clovaai/aasist/tree/main/config): `config`
- [Score output](https://github.com/clovaai/aasist/blob/main/main.py): `main.py`

### Data preparation

- **ASVspoof audio**
- [Dataset download](https://github.com/clovaai/aasist/blob/main/download_dataset.py): `download_dataset.py`
- [Trial loading](https://github.com/clovaai/aasist/blob/main/data_utils.py): `data_utils.py`

### Detection models

- [AASIST models](https://github.com/clovaai/aasist/blob/main/models/AASIST.py): `models/AASIST.py`
- [RawNet2 baseline](https://github.com/clovaai/aasist/blob/main/models/RawNet2Spoof.py): `models/RawNet2Spoof.py`
- [RawGAT-ST baseline](https://github.com/clovaai/aasist/blob/main/models/RawNetGatSpoofST.py): `models/RawNetGatSpoofST.py`
- [Pretrained weights](https://github.com/clovaai/aasist/tree/main/models/weights): `models/weights`

### Optimization and metrics

- [Training utilities](https://github.com/clovaai/aasist/blob/main/utils.py): `utils.py`
- [Evaluation metrics](https://github.com/clovaai/aasist/blob/main/evaluation.py): `evaluation.py`
- **ASV reference scores**

### Other components

- **Researcher**

## Connections

- Researcher → Experiment runner: runs (evidence: `README.md`)
- Researcher → Dataset download: requests download (evidence: `README.md`)
- Dataset download → ASVspoof audio: downloads archive (evidence: `download_dataset.py`)
- Researcher → Experiment configs: selects config (evidence: `README.md`)
- Experiment configs → Experiment runner: configures run (evidence: `README.md`)
- ASVspoof audio → Trial loading: provides audio (evidence: `data_utils.py`)
- Experiment runner → Trial loading: builds loaders (evidence: `main.py`)
- Experiment runner → AASIST models: loads architecture (evidence: `main.py`)
- Experiment runner → RawNet2 baseline: loads architecture
- Experiment runner → RawGAT-ST baseline: loads architecture
- Experiment runner → Pretrained weights: loads pretrained weights (evidence: `README.md`)
- Experiment runner → Training utilities: uses training utilities (evidence: `main.py`)
- Experiment runner → Score output: writes scores (evidence: `main.py`)
- Experiment runner → Evaluation metrics: computes metrics (evidence: `main.py`)
- ASV reference scores → Evaluation metrics: supplies ASV scores
