vijithsai/wav2vec2_cnn_ensemble

🎧 Wav2Vec2 + CNN Ensemble – Deepfake Audio Detection

Author: Vijith Sai Dataset: Hemg/DeepfakeAudio Framework: PyTorch + HuggingFace Transformers Task: Classify Real vs Fake Audio.

---

🧠 Model Architecture

• Wav2Vec2 feature embeddings
• CNN (Mel Spectrogram Encoder)
• Fusion Layer
• Final Classification Head

---

📊 Performance

• First epoch accuracy: 65%
• Fully fine-tuned ensemble model

---

🗂 Files Included

• model.safetensors
• wav2vec_state.safetensors
• feature_extractor.json
• preprocessor_config.json
• config.json
• README.md

---

🚀 Usage Example

import torch
from safetensors.torch import load_file
from transformers import AutoFeatureExtractor

# Load ensemble weights
state_dict = load_file('model.safetensors')

# Load feature extractor
feat = AutoFeatureExtractor.from_pretrained('vijithsai/wav2vec2_cnn_ensemble')