This repository contains a Jupyter Notebook (ESC-50-with-Augmentation.ipynb) that demonstrates the process of training deep learning models for audio classification using the ESC-50 dataset. The ESC-50 dataset consists of 2000 environmental audio recordings categorized into 50 different classes. The notebook covers data preprocessing, data augmentation, model building, training, evaluation, and comparison.
To run the notebook successfully, you will need to download the ESC-50 dataset from the official repository: ESC-50 Dataset. Make sure to set the dataset path variable within the code to the location where you have stored the dataset on your local machine.
The following Python libraries are used in this project:
- NumPy
- Matplotlib
- Librosa
- Scikit-learn
- TensorFlow
- Keras
- Torchaudio
- Scikit-image
Make sure to have these libraries installed in your Python environment to execute the notebook.
-
Download the ESC-50 dataset from the provided link and set the dataset path variable in the Jupyter Notebook to the location where you have stored the dataset.
-
Open the
ESC-50-with-Augmentation.ipynbnotebook in a Jupyter Notebook environment. -
Execute the notebook cells sequentially to perform the following tasks:
- Data preprocessing and feature extraction.
- Data Augmentation
- Building and training deep learning models (VGG16, VGG19, ResNet152, Ensemble model).
- Model evaluation using metrics like accuracy, log loss, precision, recall, and F1-score.
- Comparing the performance of different models.
-
Analyze the results and insights provided in the notebook to better understand Ensemble-deep learning-based audio classification.
- The ESC-50 dataset was created by Karol J. Piczak and can be found at ESC-50 Dataset.
- This project is for educational purposes and serves as a guide to working with audio data and deep learning models.
Feel free to reach out if you have any questions or need further assistance with running the notebook or understanding the code.
Happy learning!