Skip to main content

pyanthem - an audiovisualization tool to make your data more interesting

Project description

Features:

  1. Converts three-dimensional datasets to visually pleasing, audible representations

  2. A variety of video and audio parameters

  3. In-place video and audio merge

  4. GUI and Command-line interface for maximum flexibility

  5. Example datasets to get you started

  6. Headache-free installation with an Anaconda environment

Requirements

Python 3.7:

Currently, pyanthem is tested to work on Python 3.7. This will be updated as more versions are tested.

ffmpeg:

ffmpeg enables video creation and merging.

fluidsynth

fluidsynth is a powerful software synthesizer, which enables conversion of data to crisp, high quality sound files.

Usage

pyanthem was primarily developed to interpret decomposed functional imaging datasets - a dataset V with shape [height,width,time], decomposed into two lower dimensional matrixes W with shape [height*width,n], and H with shape [n,time] such that H x W = V. Here, n represents the number of variables represented by the decomposition. There are various techniques used to decompose matrixes, and it is entirely up to you how you decompose your data - two popular techniques include Non-negative Matrix Factorization (NMF), and Singular Value Decomposition (SVD).

Here’s a visual illustration of NMF - note that in this example, n=2:

https://upload.wikimedia.org/wikipedia/commons/f/f9/NMF.png

If you would prefer to keep things simple, you can skip matrix decomposition altogether and focus solely on converting raw data to audio - only working with the H matrix, where each row represents a variable and each column represents a time point. This approach only produces audio files, and it’s up to you if/how you want to merge the audio with your own visual representation.

If this is too much information to digest, don’t worry! Try the example below to get more familiar with what the data looks like, and how pyanthem transforms it.

Installation

Note: If you do not have working installations of the listed requirements (Python 3.7 + fluidsynth + ffmpeg), it is strongly recommended that you use miniconda/Anaconda for a straightforward installation process. If you do not have either, miniconda is preferred as it is a faster install and takes up much less space than Anaconda.

If you do have the above requirements installed, you can install pyanthem using pip: pip install pyanthem

Using Miniconda/Anaconda:

Create an environment and install the required packages:

conda create -n pyanthem python=3.7 pip ffmpeg fluidsynth --channel conda-forge --channel nicthib

Next, activate the environment:

conda activate pyanthem

Finally, install the pyanthem Python package using pip:

pip install pyanthem

Downloading Example datasets

If you want to get familiar with the datasets that pyanthem uses and try out the example below, download this collection of datasets/config files

(Optional) Using pyanthem in a Jupyter Notebook

To access the pyanthem environment in a Jupyter notebook, first install ipykernel in your environment:

conda install -c anaconda ipykernel

After this, create the kernel:

python -m ipykernel install --user --name=pyanthem

Once in a notebook, switch to the pyanthem kernel by selecting Kernel > Change kernel > pyanthem

Note: While the pyanthem kernel will now be available in any Jupyter notebook session, pyanthem will not function properly unless the Jupyter notebook is launched inside the pyanthem environment.

Example

Starting a pyanthem session (GUI)

First, import pyanthem and begin a pyanthem session:

import pyanthem
pyanthem.run()

The first time you run pyanthem, it will download a necessary soundfont file - this will take a minute or two.

♫ Initializing soundfont library...
♫ Downloading 1alSnxnB0JFE6mEGbUZwiYGxt2UsoO3pM into...
♫ 970.9 MiB Done.

Once completed, the pyanthem GUI will initialize:

https://github.com/nicthib/pyanthem/blob/media/GUI1.png

Next, load a dataset by clicking File > Load from .mat. Currently, you can import any .mat file that contains the following variables:

  1. H (required): A 2D matrix of shape [n,t], where each row is a component and each column is a time-point. This variable is referred to as “H” in the pyanthem environment.

  2. W (optional): A 3D matrix of shape [x,y,n], where x and y represent the spatial height and width of your dataset. If this variable is not given, no video output is possible.

  3. fr (optional): A single float value, representing the frame rate of your dataset in Hz. If a framerate is not given, pyanthem will provide a default.

If you’re having trouble, try using the example datasets linked above. For this section, we will load the dataset demo1.mat. Once loading is complete, the GUI should update with default options, and plots of H and W:

https://github.com/nicthib/pyanthem/blob/media/GUI2.png

The bottom left plots show two representations of the dataset: A preview of the output movie (left), and a visualization of what components are included and the colormap selection. The right two plots show raw representations of H (top), and a visualization of the audio output file (right). Lighter colors indicate loud notes, and darker colors indicate quiet notes, with black indicating silence.

From here, you can adjust parameters, preview the output, and finally save video and audio files. If you want to check how your parameter adjustments impact your audivisualization, click the Update button, and your changes will be reflected. Any issues with your selected parameters will be indicated in the white status box. Try adjusting a few parameters and observing how the plots change.

Finally, render output files with the Save menu.

Using pyanthem in CLI (command-line interface) mode

pyanthem’s CLI mode is useful for running batch conversions of large amounts of data once you are happy with your audiovisualization parameters, and isn’t necessary until you have used the GUI and would like to automate your conversions.

To run pyanthem in CLI mode, pass the argument display=False, and assign the .run() method to a variable:

import pyanthem
g=pyanthem.run(display=False)

Next, load a dataset and config file using the .load_data() and .load_config() methods. You can pass an explicit file name to the file_in argument, or pass none to recieve a file select prompt (note the use of the leading r when naming a file location):

g.load_data(file_in=r'path/to/your/file.mat')
g.load_config(file_in=r'path/to/your/config.p')

Finally, render the audio and videofiles, then merge the outputs using the .write_audio(), .write_video() and .merge() methods:

g.write_audio()
g.write_video()
g.merge()

Once you’re comfortable with this syntax, you can combine all of these steps into a single line, write a merged video with the .write_AV() method, and even remove the intermediate files using the .cleanup() method:

data_file = r'path/to/your/file.mat'
config_file = r'path/to/your/config.p'
g.load_data(file_in=data_file).load_config(file_in=config_file).write_AV().cleanup()

Team

Nic Thibodeaux

niclogo

http://github.com/nicthib

FAQ

Under construction!

Support

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pyanthem-1.1.2.tar.gz (20.5 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page