DAT255 Deep Learning project

Google Speech Command dataset

The dataset used for training, validating and testing the deep neural network, is formatted like the following:

• file: Path to the audio file.
• audio['array']: The actual audio data as a numpy array.
• audio['sampling_rate']: The sampling rate of the audio file (16KHz).
• label: The label for the audio file (e.g. a specific command like "yes", "no", etc.).
• is_unknown: Boolean flag indicating wether the sample is unknown.
• speaker_id: The speakers ID who recordede the sample.
• utterance_id: The ID of the specific utterance of that speaker.

---

Installation and Usage

Follow these steps to set up the project locally:

1. Clone the repository:

   git clone https://github.com/petterr-n/255-Project.git
   

2. Navigate to the project directory:

   cd your-repository-name
   

3. Set up a virtual environment:

   python3 -m venv venv
   

4. Activate the virtual environment:

   • On macOS/Linux:

     source venv/bin/activate
     

5. Install dependencies:

   pip install -r requirements.txt
   

---

Usage

To run the project:

1. Activate the virtual environment (if not already active) (Steg 4).

2. Run the main script:

   python main.py
   

---