Beijuka/voice-gender-classifier

Usage

import torch
from model import ECAPA_gender
# You could directly download the model from the huggingface model hub
model = ECAPA_gender.from_pretrained("JaesungHuh/voice-gender-classifier")
model.eval()
# If you are using gpu .... 
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
# Load the audio file and use predict function to directly get the output
example_file = "data/00001.wav"
with torch.no_grad():
    output = model.predict(example_file, device=device)
    print("Gender : ", output)

Pretrained weights

For those who need pretrained weights, please download it in here

Training details

State-of-the-art speaker verification model already produces good representation of the speaker's gender.

I used the pretrained ECAPA-TDNN from TaoRuijie's repository, added one linear layer to make two-class classifier, and finetuned the model with the VoxCeleb2 dev set.

The model achieved 98.7% accuracy on the VoxCeleb1 identification test split.

Caveat

I would like to note the training dataset I've used for this model (VoxCeleb) may not represent the global human population. Please be careful of unintended biases when using this model.

Reference