Context Recognition

A deep learning project for environmental sound classification using convolutional neural networks.

Overview

This project implements a sound classification system capable of recognizing different environmental contexts based on audio data. It uses the ESC-50 dataset (Environmental Sound Classification) and implements convolutional neural networks with PyLearn2 and Theano.

Dataset

The project uses the ESC-50 dataset, a labeled collection of 2000 environmental audio recordings suitable for benchmarking methods of environmental sound classification.

Project Structure

• Dataset.ipynb: Data processing utilities and dataset preparation
• Net-DoubleConv.ipynb: Implementation of the convolutional neural network architecture
• EDA on ESC50 Dataset.ipynb: Exploratory data analysis on the ESC-50 dataset
• Evaluation.ipynb: Model evaluation utilities and metrics

Requirements

The project requires:

• Python 2.7
• PyLearn2
• Theano
• NumPy
• Pandas
• scikit-learn
• IPython/Jupyter

How to Use

1. Clone the repository
2. Download the ESC-50 dataset
3. Run the notebooks in the following order:
   • First, explore the dataset using EDA on ESC50 Dataset.ipynb
   • Prepare the dataset using Dataset.ipynb
   • Train the model with Net-DoubleConv.ipynb
   • Evaluate the model using Evaluation.ipynb

Model Architecture

The model uses a double convolutional neural network architecture with:

• Multiple convolutional layers with ReLU activation
• Max pooling layers
• Dropout for regularization
• Softmax output layer for classification

Results

The model is evaluated on sound classification accuracy across different environmental contexts.