Eq-Detection

Installation

The package uses poetry as a dependency manager, so you need to install it first. You can find the installation instructions here.

Then install the package dependencies and executables using make: bash make install

Note that project dependencies are splitted in docs, dev, notebooks and the required ones for the package to work, by default make install will install everithing except the notebooks dependencies, if you want to install all in once you can use poetry install.

Docker

To avoid any issue with setup and installations a docker container is provided launch the command to create a local container called mpai: bash docker build -t mpai .

Usage

The package come with three executables:

• noise-extractor: Extracts noise from audio files and saves it in a folder
• noise-classifier: Classifies noise files using a pre-trained model and generates a report in form of Irregularity files
• audio-analyser: executes the whole audio analysis pipeline, extracting noise, classifying it and generating the irregularity files

To run the commands from the container use the command: bash docker run -it mpai /bin/bash and then you can run the commands as described below.

Since the container main use is to execute CLI commands the scripts folder contains some wrapper scripts to execute the CLI commands from outside the container, to execute the noise-extractor command type the command: bash ./scripts/docker-noise-extractor.py -h

noise-extractor

Once installed you can run the package using the command from the poetry shell[^1] ```bash noise-extractor -h

or poetry run noise-extractor -h if you are not in the poetry shell

This will output the help message: bash usage: noise-extractor [-h] -i INPUT -d DESTINATIONDIR -l MINSILENCE_LEN [-q]

A tool to extract noise from sound files

options: -h, --help show this help message and exit -i INPUT, --input INPUT Specify a single file or a folder containing audio files as input -d DESTINATIONDIR, --destination-dir DESTINATIONDIR Where to put file chunks -l SILENCELEN, --silence-len SILENCELEN The duration in milliseconds of silence to detect -q, --quiet Display less output (errors only) ```

noise-classifier

```bash noise-classifier -h

or poetry run noise-classifier -h if you are not in the poetry shell

```

audio-analyser

```bash audio-analyser -h

or poetry run audio-analyser -h if you are not in the poetry shell

```

Docs

A more detailed documentation can be found in the docs folder. This docs can be rendered to HTML or PDF using sphinx. To read the docs in HTML format type the command make docs from the root folder, it will open the system browser on the generated documentation, while make docs-pdf will do the same as make docs but generates a pdf otput. Otherwise you can run the Makefile in the docs folder and type make html or make latexpdf to generate the documentation in PDF format (make help will output a list of all possible formats), the generated files will be in the docs/build folder.

The first time you run the make docs command it will take a while to generate the documentation, because it generates the images and code examples by executing some python scripts, but after the first rendering it will be much faster.

Notebooks

To execute the notebooks you can open them in your favourite way (using conda or similar environments) or install the notebooks dependencies and run them using jupyter from the poetry virtual environment. To install the notebooks dependencies type the command: bash make install-notebooks Then you can run the notebooks using the command: bash poetry run jupyter notebook ./notebooks

[^1]: If you are not familiar with poetry, you can find more information here.