What’s the Task?

You will work with audio files and build a basic machine-learning model. The goal is to clean up raw audio data, turn it into something a computer can learn from, and then train a simple model to show that it works. Think of it as laying the groundwork for a smarter audio-based system.

What You’ll Do:

1. Preprocess the Audio Data

• Load and Parse Audio Files:

  Use tools like librosa and soundfile to read and work with the raw audio files.

• Extract Useful Features:

  Pull out important audio characteristics like:

  • MFCCs (Mel-frequency cepstral coefficients)
  • Chroma features
  • Spectral contrast

• Clean the Data:

  Fix any problems in the data (e.g., missing values, noise, inconsistencies).

• Build the Dataset:

  Organize the cleaned and processed data into a pandas DataFrame, and ensure everything is labeled and ready for training (including splitting into train/test sets).

2. Train a Simple Model

• Build a Baseline Model:

  Use Python libraries like scikit-learn (or similar) to create a basic model.

• Train and Evaluate:

  Train your model on the data and share how well it performs (accuracy, loss, etc.).

3. Share Your Work

• Provide a clean and commented Python script or Jupyter Notebook that shows everything you did.
• Save the final processed dataset in a shareable format (like CSV or pickle).
• Write a short report that explains:
  • What steps you took
  • Any challenges you faced
  • How your model performed
  • Suggestions for what could be improved or done next

⏳ Deadline