|
|
--- |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
base_model: |
|
|
- facebook/wav2vec2-base-960h |
|
|
pipeline_tag: audio-classification |
|
|
library_name: transformers |
|
|
tags: |
|
|
- voice-gender-detection |
|
|
- male |
|
|
- female |
|
|
- biology |
|
|
- SFT |
|
|
--- |
|
|
|
|
|
 |
|
|
|
|
|
# Common-Voice-Gender-Detection |
|
|
|
|
|
> **Common-Voice-Gender-Detection** is a fine-tuned version of `facebook/wav2vec2-base-960h` for **binary audio classification**, specifically trained to detect speaker gender as **female** or **male**. This model leverages the `Wav2Vec2ForSequenceClassification` architecture for efficient and accurate voice-based gender classification. |
|
|
|
|
|
> [!note] |
|
|
Wav2Vec2: Self-Supervised Learning for Speech Recognition : [https://arxiv.org/pdf/2006.11477](https://arxiv.org/pdf/2006.11477) |
|
|
|
|
|
```py |
|
|
Classification Report: |
|
|
|
|
|
precision recall f1-score support |
|
|
|
|
|
female 0.9705 0.9916 0.9809 2622 |
|
|
male 0.9943 0.9799 0.9870 3923 |
|
|
|
|
|
accuracy 0.9846 6545 |
|
|
macro avg 0.9824 0.9857 0.9840 6545 |
|
|
weighted avg 0.9848 0.9846 0.9846 6545 |
|
|
``` |
|
|
|
|
|
 |
|
|
|
|
|
 |
|
|
|
|
|
--- |
|
|
|
|
|
## Label Space: 2 Classes |
|
|
|
|
|
``` |
|
|
Class 0: female |
|
|
Class 1: male |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Install Dependencies |
|
|
|
|
|
```bash |
|
|
pip install gradio transformers torch librosa hf_xet |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Inference Code |
|
|
|
|
|
```python |
|
|
import gradio as gr |
|
|
from transformers import Wav2Vec2ForSequenceClassification, Wav2Vec2FeatureExtractor |
|
|
import torch |
|
|
import librosa |
|
|
|
|
|
# Load model and processor |
|
|
model_name = "prithivMLmods/Common-Voice-Geneder-Detection" |
|
|
model = Wav2Vec2ForSequenceClassification.from_pretrained(model_name) |
|
|
processor = Wav2Vec2FeatureExtractor.from_pretrained(model_name) |
|
|
|
|
|
# Label mapping |
|
|
id2label = { |
|
|
"0": "female", |
|
|
"1": "male" |
|
|
} |
|
|
|
|
|
def classify_audio(audio_path): |
|
|
# Load and resample audio to 16kHz |
|
|
speech, sample_rate = librosa.load(audio_path, sr=16000) |
|
|
|
|
|
# Process audio |
|
|
inputs = processor( |
|
|
speech, |
|
|
sampling_rate=sample_rate, |
|
|
return_tensors="pt", |
|
|
padding=True |
|
|
) |
|
|
|
|
|
with torch.no_grad(): |
|
|
outputs = model(**inputs) |
|
|
logits = outputs.logits |
|
|
probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist() |
|
|
|
|
|
prediction = { |
|
|
id2label[str(i)]: round(probs[i], 3) for i in range(len(probs)) |
|
|
} |
|
|
|
|
|
return prediction |
|
|
|
|
|
# Gradio Interface |
|
|
iface = gr.Interface( |
|
|
fn=classify_audio, |
|
|
inputs=gr.Audio(type="filepath", label="Upload Audio (WAV, MP3, etc.)"), |
|
|
outputs=gr.Label(num_top_classes=2, label="Gender Classification"), |
|
|
title="Common Voice Gender Detection", |
|
|
description="Upload an audio clip to classify the speaker's gender as female or male." |
|
|
) |
|
|
|
|
|
if __name__ == "__main__": |
|
|
iface.launch() |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## Demo Inference |
|
|
|
|
|
> [!note] |
|
|
male |
|
|
|
|
|
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/7woMf3_bgX_D99-1Uy3jH.mpga"></audio> |
|
|
|
|
|
 |
|
|
|
|
|
|
|
|
> [!note] |
|
|
female |
|
|
|
|
|
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/0d2rDf_DT-gjRWBwiPbm_.mpga"></audio> |
|
|
|
|
|
 |
|
|
|
|
|
--- |
|
|
|
|
|
## Intended Use |
|
|
|
|
|
`Common-Voice-Gender-Detection` is designed for: |
|
|
|
|
|
* **Speech Analytics** – Assist in analyzing speaker demographics in call centers or customer service recordings. |
|
|
* **Conversational AI Personalization** – Adjust tone or dialogue based on gender detection for more personalized voice assistants. |
|
|
* **Voice Dataset Curation** – Automatically tag or filter voice datasets by speaker gender for better dataset management. |
|
|
* **Research Applications** – Enable linguistic and acoustic research involving gender-specific speech patterns. |
|
|
* **Multimedia Content Tagging** – Automate metadata generation for gender identification in podcasts, interviews, or video content. |
