README.md

Deep Learning Project Sound Classification

Prerequisites

To run this project, you need to have Anaconda installed. You can download and install Anaconda from here.

Download the needed features from here. (see Directory structure)

Download UrbanSound8k dataset from here and extract it the folder UrbanSound(see Directory structure)

Setting Up the Environment

1. Create a Conda environment with Python 3.10:

   conda create -n env python=3.10

2. Activate the Conda environment:

   conda activate env

3. Install the required packages:

   pip install -r requirements.txt

Running the Notebooks

Carefully read through the Jupyter notebooks and run them in order. The sequence is important for ensuring that all necessary steps are completed:

1. 1.Urban_Sound_Pipeline
2. 2.Audioset_Pipeline
3. 3.Transfer_Learning_Pipeline
4. '4. ESC50_Pipeline'

4.Demo is intended for demonstration purposes.

Directory Structure

If you download features and models from Google Drive, ensure that the directory structure is arranged as follows:

Project
├── Audioset
│   ├── AS_Eval (Only if you download the audio)
│   ├── AS_Train (Only if you download the audio)
│   ├── CSVs
│   ├── Train_Features
│   └── Eval_Features
├── Bonus
├── Models
├── Notebooks
├── Scripts
└── UrbanSound
    ├── CSVs
    ├── Features
    │   ├── CNN_Features
    │   ├── NN_Features
    │   └── YamNet_Features
    ├── Test
    ├── Train
    └── UrbanSound8K (Original Folder)
└── ESC500
    ├── CSVs
    ├── Features
    │   ├── CNN_Features  
    ├── Test
    ├── Train
    └── ESC-50-master (Original Folder)