Skip to content

Latest commit

 

History

History
77 lines (56 loc) · 3.12 KB

README.md

File metadata and controls

77 lines (56 loc) · 3.12 KB
title emoji colorFrom colorTo sdk sdk_version app_file pinned license short_description
Gender Detection
🚀
gray
indigo
streamlit
1.42.0
app.py
false
apache-2.0
real time gender detection app`

Real-Time Gender detection app

This is a real-time gender detection app that uses a pre-trained wav2vec2 model to classify the gender of a speaker from their voice. The app is built using Streamlit and leverages the power of transformers from the Hugging Face library.

Demo Video

Watch the demo video below to see the app in action:

Real-Time Gender Detection Demo

Features

  • Real-time audio processing
  • Gender classification using a pre-trained wav2vec2 model
  • Simple and intuitive user interface

How to Use

  1. Clone the repository to your local machine.
  2. Ensure you have Python 3.10 installed on your system.
  3. Install the required dependencies using pip install -r requirements.txt.
  4. Run the app using streamlit run app.py.
  5. Click the 'Start' button to begin real-time gender detection from your microphone input.

Dependencies

  • streamlit
  • numpy
  • torch
  • transformers
  • pyaudio

Model

The app uses the alefiury/wav2vec2-large-xlsr-53-gender-recognition-librispeech model from Hugging Face for gender recognition.

Approach

The app uses the alefiury/wav2vec2-large-xlsr-53-gender-recognition-librispeech model from Hugging Face for gender recognition. The model is a pre-trained wav2vec2 model for gender recognition on the LibriSpeech dataset. The model is loaded using the AutoFeatureExtractor and AutoModelForAudioClassification classes from the Hugging Face library.

The app uses the pyaudio library to capture the audio from the microphone. The audio is captured using the pyaudio library and the AudioStream class. The audio is captured using the AudioStream class and the start_stream method. The audio is captured using the AudioStream class and the stop_stream method.

The app uses the streamlit library to create the web application. The app is configured to use a 16kHz sampling rate and mono audio input.

Configuration

The app is configured to use a 16kHz sampling rate and mono audio input. The audio stream parameters are defined as follows:

  • FORMAT: 16-bit resolution
  • CHANNELS: 1 (Mono audio)
  • RATE: 16000 (16kHz sampling rate)
  • CHUNK: 1024 (Number of frames per buffer)

Logging

The app uses Python's built-in logging module to log information with a basic configuration that includes timestamps and log levels.

License

This project is licensed under the Apache-2.0 License.

Acknowledgements

  • Hugging Face for providing the pre-trained wav2vec2 model.
  • Streamlit for the easy-to-use web application framework.
  • PyAudio for handling audio input.

Huggingface Space deployment

https://huggingface.co/spaces/prashant-garg/gender-detection

Note: The model is hosted on Huggingface and the app is deployed on Huggingface Spaces. But there is an issue with the audio input. The app is not able to capture the audio from the microphone as huggingface spaces does not have audio drivers enabled.