Loading
Du hast einen ComfyUI-Workflow? Erstelle direkt im Browser eine verständliche Workflow-Dokumentation.Workflow verständlich machen.Jetzt ausprobieren ↗
Reproduction of "Audio Deepfake Detection with XLS-R and SLS Classifier" (Zhang et al., ACM Multimedia 2024).
Modellquelle
Quellenbeschreibung
Reproduction of "Audio Deepfake Detection with XLS-R and SLS Classifier" (Zhang et al., ACM Multimedia 2024).
The Selective Layer Summarization (SLS) classifier extracts attention-weighted features from all 24 transformer layers of XLS-R 300M (wav2vec 2.0), then classifies bonafide vs. spoofed speech via a lightweight fully-connected head. RawBoost (algo=3, SSI) data augmentation is applied during training.
Quellen
1 QuelleVerifiziert 13. Sept.
Modellartefakte
2 ArtefakteQuellenauszüge
2 Auszüge| File | Experiment | Description |
|---|---|---|
v1/epoch_2.pth | v1 (baseline) | Best cross-domain generalization. Patience=1, no validation, 4 epochs. |
v2/epoch_16.pth | v2 (val-based) | Validation early stopping. Patience=10, ASVspoof2019 LA dev validation, 27 epochs. |
Recommended: Use v1/epoch_2.pth — it generalizes better to unseen attack types (DF, In-the-Wild).
The original pretrained checkpoints from Zhang et al. are available from:
| Track | Paper EER (%) | v1 EER (%) | v2 EER (%) |
|---|---|---|---|
| ASVspoof 2021 DF | 1.92 | 2.14 | 3.75 |
| ASVspoof 2021 LA | 2.87 | 3.51 | 3.47 |
| In-the-Wild | 7.46 | 7.84 | 12.67 |
v1 closely reproduces the paper results. v2 improves LA slightly but degrades DF and In-the-Wild due to overfitting to the LA validation domain — a well-documented cross-domain generalization problem in audio deepfake detection (Muller et al., Interspeech 2022).
Both experiments share the following setup:
| Parameter | Value |
|---|---|
| Training data | ASVspoof2019 LA train (25,380 utterances) |
| Loss | Weighted Cross-Entropy [0.1, 0.9] |
| Optimizer | Adam (lr=1e-6, weight_decay=1e-4) |
| Batch size | 5 |
| RawBoost | algo=3 (SSI) |
| Seed | 1234 |
| SSL backbone | XLS-R 300M (frozen feature extractor) |
| GPU | NVIDIA RTX 4080 (16 GB) |
torch.no_grad() in validation loop, correct best_val_loss trackingfrom huggingface_hub import hf_hub_download
# Download v1 checkpoint (recommended)
checkpoint_path = hf_hub_download(
repo_id="sukhdeveyash/XLS-R-SLS-Deepfake-Detection",
filename="v1/epoch_2.pth"
)
# Download v2 checkpoint
# checkpoint_path = hf_hub_download(
# repo_id="sukhdeveyash/XLS-R-SLS-Deepfake-Detection",
# filename="v2/epoch_16.pth"
# )
import torch
from model import Model # from the GitHub repo
device = "cuda" if torch.cuda.is_available() else "cpu"
model = Model(device=device, ssl_cpkt_path="xlsr2_300m.pt")
model.load_state_dict(torch.load(checkpoint_path, map_location=device))
model = model.to(device)
model.eval()
Full training and evaluation code: GitHub Repository
xlsr2_300m.pt) from fairseqSee environment.yml in the GitHub repo for the full environment.
@inproceedings{zhang2024audio,
title={Audio Deepfake Detection with XLS-R and SLS Classifier},
author={Zhang, Qishan and Wen, Shuangbing and Hu, Tao},
booktitle={Proceedings of the 32nd ACM International Conference on Multimedia},
year={2024},
publisher={ACM}
}
--- license: mit tags: - audio - deepfake-detection - anti-spoofing - wav2vec2 - xlsr - speech - asvspoof datasets: - asvspoof2019 - asvspoof2021 metrics: - equal_error_rate pipeline_tag: audio-classification language: - en library_name: pytorch --- # XLS-R + SLS Classifier for Audio Deepfake Detection Reproduction of **"Audio Deepfake Detection with XLS-R and SLS Classifier"** (Zhang et al., ACM Multimedia 2024). The Selective Layer Summarization (SLS) classifier extracts attention-weighted features from all 24 transformer layers of [XLS-R 300M](https://huggingface.co/facebook/wav2vec2-xls-r-300m) (wav2vec 2.0), then classifies bonafide vs. spoofed speech via a lightweight fully-connected head. [RawBoost](https://arxiv.org/abs/2301.00693) (algo=3, SSI) data augmentation is applied during training. ## Available Checkpoints | File | Experiment | Description | |------|-----------|-------------| | `v1/epoch_2.pth` | v1 (baseline) | Best cross-domain generalization. Patience=1, no validation, 4 epochs. | | `v2/epoch_16.pth` | v2 (val-based) | Validation early stopping. Patience=10, ASVspoof2019 LA dev validation, 27 epochs. | **Recommended**: Use `v1/epoch_2.pth` — it generalizes better to unseen attack types (DF, In-the-Wild). ### Original authors' pretrained models The original pretrained checkpoints from Zhang et al. are available from: - [Google Drive](https://drive.google.com/drive/folders/13vw_AX1jHdYndRu1edlgpdNJpCX8OnrH?usp=sharing) - [Baidu Pan](https://pan.baidu.com/s/1dj-hjvf3fFPIYdtHWqtCmg?pwd=shan) (password: shan) ## Results | Track | Paper EER (%) | v1 EER (%) | v2 EER (%) | |-------|--------------|------------|------------| | ASVspoof 2021 DF | 1.92 | **2.14** | 3.75 | | ASVspoof 2021 LA | 2.87 | 3.51 | **3.47** | | In-the-Wild | 7.46 | **7.84** | 12.67 | v1 closely reproduces the paper results. v2 improves LA slightly but degrades DF and In-the-Wild due to overfitting to the LA validation domain — a well-documented cross-domain generalization problem in audio deepfake detection ([Muller et al., Interspeech 2022](https://arxiv.org/abs/2203.16263)). ## Training Configuration Both experiments share the following setup: | Parameter | Value | |-----------|-------| | Training data | ASVspoof2019 LA train (25,380 utterances) | | Loss | Weighted Cross-Entropy [0.1, 0.9] | | Optimizer | Adam (lr=1e-6, weight_decay=1e-4) | | Ba...
Source context: 0 downloads · 0 likes · Pipeline audio-classification · Library pytorch · Repo sukhdeveyash/XLS-R-SLS-Deepfake-Detection