W

Wav2vec2 Base 960h Asv19 Deepfake

Developed by abhishtagatya
An audio classification model fine-tuned based on facebook/wav2vec2-base-960h, specializing in deepfake and audio spoofing detection.
Downloads 25
Release Time : 3/12/2024

Model Overview

This model is used to detect deepfake and spoofing behaviors in audio, demonstrating high accuracy and low error rates on evaluation datasets.

Model Features

High Accuracy
Achieves 98.45% accuracy on evaluation datasets
Low Error Rate
False Acceptance Rate (FAR) 0.9%, False Rejection Rate (FRR) 1.62%, Equal Error Rate (EER) 1.26%
Based on Proven Architecture
Fine-tuned on the validated wav2vec2-base-960h architecture

Model Capabilities

Audio Classification
Deepfake Detection
Audio Spoofing Detection

Use Cases

Security Verification
Voice Authentication
Detects forged audio in voice authentication systems
Effectively identifies 98.45% of forged audio
Content Moderation
Audio Content Moderation
Identifies tampered or forged audio content
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase