🎙️ Accent Classification using Wav2Vec2

A Deep Learning project that uses HuggingFace's Wav2Vec2 model to classify English audio accents as Canadian 🇨🇦 or British (England) 🏴.

Built with PyTorch, Transformers, and Librosa, this project achieves high accuracy on voice-based accent prediction.

---

---

✅ Results

---

🤗 Try the Live Demo on 👉 Hugging Face Spaces

🧠 Live Accent Classification App deployed on Hugging Face !!

---

📊 Model Performance Visualizations

Below are various graphs and charts generated to evaluate and visualize the performance of the Wav2Vec2-based Accent Classifier:

🔹 Horizontal Bar Chart

Class-wise prediction confidence of the model.

---

🔹 Training Loss Curve

Loss reduction over training epochs showing model learning progress.

---

🔹 Confusion Matrix

Matrix showing the number of correct and incorrect predictions per class.

---

🖥️ Gradio Web App

A sleek, dark-themed Gradio interface has been added to make accent classification interactive and user-friendly!

• 🎤 Supports both microphone recording and file upload.
• 🧠 Uses the fine-tuned Wav2Vec2 model for detecting Canadian or British English accents.
• 🌌 Clean dark mode layout with colorful buttons for a smooth user experience.
• ✅ Built with gradio==5.38.2.

🚀 Simply run app.py to launch the app locally:

python app.py

🔹 Gradio interface Before Prediction

🔹 Gradio interface After Prediction

🚀 Features

• Fine-tuned Wav2Vec2-Large-960h model
• Custom classification head for binary accent classification
• Achieved 97.83% Accuracy, 100% Precision, 97.62% F1 Score
• Audio pre-processing with Librosa & Torchaudio
• Exportable results in CSV + visualizations
• Easy-to-use Jupyter Notebook interface

---

🛠️ Tech Stack

• Python
• PyTorch
• HuggingFace Transformers
• Librosa
• Scikit-learn
• Matplotlib / Seaborn
• Git / GitHub

---

🧰 Setup Instructions

Follow these steps to set up the project on your local machine.

1️⃣ Prerequisites

• Python 3.9 or higher
• Git installed (Download Git)

2️⃣ Clone the Repository

git clone https://github.com/creativepurus/Accent_Classification.git
cd Accent_Classification

3️⃣ Create & Activate a Virtual Environment

# Create venv (do this only once)
python -m venv venv

# Activate venv
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

4️⃣ Install Dependencies

pip install -r requirements.txt

5️⃣ Run the Jupyter Notebook

jupyter notebook

Open en_model.ipynb and run the cells step-by-step.

---

📊 Outputs

• model_predictions.csv → Stores predictions with confidence
• Graphs folder → Contains plots like confusion matrix, class-wise distribution
• Trained model saved to wav2vec2-accent-classifier/

---

🗂️ Project Structure

Accent_Classification/
├── .github/                      # GitHub workflows, actions, or templates
├── assets/                      # Images or media assets for README or UI
├── Data/                        # Audio + tsv files
├── Graphs/                      # Evaluation plots
├── Model/                       # Saved Model and processor files (auto-created after training)
├── Accent_Detection_Model.ipynb # Main notebook
├── app.py                       # Gradio-based UI Application
├── Code_Explanation.md          # Detailed explanation of model and code
├── LICENSE                      # Project license file
├── README.md                    # Project overview and instructions
├── requirements.txt             # Python dependencies
└── .gitignore                   # Git ignore rules for GitHub

---

⚠️ Notes

• Don’t forget to activate your virtual environment before installing packages or running the notebook.
• Make sure audio files and validated.tsv are placed correctly in the en/ folder.

---

🧪 Future Improvements

• Add support for more accents (e.g., Australian, Indian, American)
• Deploy as a full-fledged web app
• Integrate with Flask/FastAPI for production API

---

👨‍💻 Author

Made with ❤️ by Anand Purushottam

LinkedIn | GitHub

Love My Work ? You can 👉🏻

---

📄 License

This project is licensed under the GNU GENERAL PUBLIC LICENSE.