🎭 EmoFake Detection App

This is a Streamlit-based web application that uses advanced machine learning models to detect whether an audio file is bonafide (emotionally authentic) or spoofed (emotionally fake). The application leverages Facebook's Wav2Vec2 for feature extraction and a TensorFlow-based CNN model trained on PTMs features for classification.

---

📋 Features

• Upload .wav audio files for classification.
• Provides predictions (bonafide or spoof) with confidence scores.
• Displays class probabilities with a visually appealing bar chart.
• Easy-to-use interface built with Streamlit.

---

🛠️ Setup Instructions

1. Clone the Repository

git clone https://github.com/gir-ish/Emo-FAKE.git
cd Emo-FAKE

2. Create a Virtual Environment (Optional but Recommended)

python -m venv env
source env/bin/activate  

3. Install Dependencies

Install the required Python libraries using the requirements.txt file:

pip install -r requirements.txt

4. Run the Application

Navigate to the APP folder and launch the Streamlit app:

cd APP
streamlit run app.py

5. If Model Doesn't Work, Download from Here

If the model isn't available in your project, you can download it from the following link:

Download Saved CNN MMS Model

After downloading: replace the model folder from new one.

---

---

🔗 How to Use

1. Open the App: After running the above command, Streamlit will provide a local URL (e.g., http://localhost:----). Open it in your browser.

2. Upload an Audio File:

   • Click on the "Browse files" button or drag and drop a .wav file.
   • Ensure the file is in WAV format and meets the required specifications (e.g., 16kHz sampling rate).

3. View Results:

   • The app will process the audio and display the following:
     • Prediction: Whether the file is "bonafide" or "spoof."
     • Confidence Score: How confident the model is in its prediction.
     • Class Probabilities: Probabilities for each class (visualized as a bar chart).

---

🖥️ Online Application

You can also use the application directly via the following link:

Access the EmoFake Detection App on Hugging Face Spaces

---

🛠️ Troubleshooting

Common Issues

1. Dependency Errors:

   • Ensure you have the correct versions of torch, transformers, tensorflow, and torchaudio. Use the provided requirements.txt to avoid version mismatches.

2. File Format Errors:

   • Ensure the uploaded file is in .wav format and is not corrupted.

3. GPU Not Detected:

   • If using a GPU, ensure CUDA is installed and the torch version matches your CUDA version.

4. Permission Issues (Windows):

   • Run Command Prompt as Administrator to resolve file permission issues.

---

📂 Project Structure

.
├── APP                  # Folder containing Streamlit application
│   ├── app.py           # Main Streamlit application
├── CNN-MODEL  # Trained TensorFlow CNN model directory
├── requirements.txt     # Python dependencies
├── README.md            # Project documentation

---

📜 License

This project is licensed under the MIT License. See the LICENSE file for details.

---