Audio Classification
Transformers
TensorBoard
Safetensors
audio-spectrogram-transformer
Generated from Trainer
Eval Results (legacy)
Instructions to use MattyB95/AST-ASVspoof5-Synthetic-Voice-Detection with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use MattyB95/AST-ASVspoof5-Synthetic-Voice-Detection with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("audio-classification", model="MattyB95/AST-ASVspoof5-Synthetic-Voice-Detection")# Load model directly from transformers import AutoFeatureExtractor, AutoModelForAudioClassification extractor = AutoFeatureExtractor.from_pretrained("MattyB95/AST-ASVspoof5-Synthetic-Voice-Detection") model = AutoModelForAudioClassification.from_pretrained("MattyB95/AST-ASVspoof5-Synthetic-Voice-Detection") - Notebooks
- Google Colab
- Kaggle
Threshold for AST MODEL
#1
by yyw2683 - opened
Hi, I'm trying to replicate the results, can you please share the threshold using for classification at your convenience?
Apologies for this delayed and slow reply.
We didn’t use a separately tuned threshold. The AST fine-tuned classifier outputs 2-class logits, and we take the argmax over the logits to obtain the predicted label (default decision rule). For the binary case, this is equivalent to using a 0.5 threshold on the softmax probability for the “Synthetic” class. I hope this helps.