Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/axegon/spectrust
Spectrust - blazing fast spectrogram generator for Python in Rust.
https://github.com/axegon/spectrust
python3 rust
Last synced: 2 days ago
JSON representation
Spectrust - blazing fast spectrogram generator for Python in Rust.
- Host: GitHub
- URL: https://github.com/axegon/spectrust
- Owner: axegon
- License: mit
- Created: 2018-11-09T15:57:30.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2020-04-20T10:01:37.000Z (over 4 years ago)
- Last Synced: 2024-11-02T03:32:20.490Z (11 days ago)
- Topics: python3, rust
- Language: Python
- Size: 4.31 MB
- Stars: 24
- Watchers: 5
- Forks: 5
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- Changelog: HISTORY.md
Awesome Lists containing this project
- awesome-blazingly-fast - spectrust - Spectrust - blazing fast spectrogram generator for Python in Rust. (Python)
README
## Spectrust - blazing fast mel-frequency spectrogram generator for Python in Rust.
```
██████ ██▓███ ▓█████ ▄████▄ ▄▄▄█████▓ ██▀███ █ ██ ██████ ▄▄▄█████▓
▒██ ▒ ▓██░ ██▒▓█ ▀ ▒██▀ ▀█ ▓ ██▒ ▓▒▓██ ▒ ██▒ ██ ▓██▒▒██ ▒ ▓ ██▒ ▓▒
░ ▓██▄ ▓██░ ██▓▒▒███ ▒▓█ ▄ ▒ ▓██░ ▒░▓██ ░▄█ ▒▓██ ▒██░░ ▓██▄ ▒ ▓██░ ▒░
▒ ██▒▒██▄█▓▒ ▒▒▓█ ▄ ▒▓▓▄ ▄██▒░ ▓██▓ ░ ▒██▀▀█▄ ▓▓█ ░██░ ▒ ██▒░ ▓██▓ ░
▒██████▒▒▒██▒ ░ ░░▒████▒▒ ▓███▀ ░ ▒██▒ ░ ░██▓ ▒██▒▒▒█████▓ ▒██████▒▒ ▒██▒ ░
▒ ▒▓▒ ▒ ░▒▓▒░ ░ ░░░ ▒░ ░░ ░▒ ▒ ░ ▒ ░░ ░ ▒▓ ░▒▓░░▒▓▒ ▒ ▒ ▒ ▒▓▒ ▒ ░ ▒ ░░
░ ░▒ ░ ░░▒ ░ ░ ░ ░ ░ ▒ ░ ░▒ ░ ▒░░░▒░ ░ ░ ░ ░▒ ░ ░ ░
░ ░ ░ ░░ ░ ░ ░ ░░ ░ ░░░ ░ ░ ░ ░ ░ ░
░ ░ ░░ ░ ░ ░ ░
```
### Motivation
This package was originally made as a supplement to a tensorflow application.
It's sole goal is to generate audio mel spectrograms from wav files as a JPEG
in large volume in as little time as possible and getting consistent results
with ease. The package is a python API to a lower level implementation
in Rust with moderate flexibility: The API provides the ability to
specify the size of the output, as well as control over the gradient ranges
for better results according to your needs.
### Installation
Package is now available in PyPI
```
pip3 install spectrust
```
If you want to compile from source, you will need to have a running Rust nightly installation,
please visit [https://www.rust-lang.org/en-US/install.html](https://www.rust-lang.org/en-US/install.html).
Once you have installed rust run:
```
pip3 install git+https://github.com/axegon/spectrust.git
```
### Usage
The API is fairly straightforward:
```python
>>> import spectrust
>>> spect = spectrust.Spectrogram(width=800, height=600)
>>> spect.generate('/home/alex/Downloads/step-5000-wav.wav', '/tmp/hello_world.jpg')
```
Generating the following output:
In addition you can pass `r`, `g` and `b` arguments to the `Spectrogram` constructor,
ranging from 0 to 255 each to achieve different results that may be more desirable
in your case. Keep in mind those values **must** be floats:
```python
>>> import spectrust
>>> spect = spectrust.Spectrogram(width=800, height=600, r=255., g=255., b=150.)
>>> spect.generate('/home/alex/Downloads/step-5000-wav.wav', '/tmp/hello_world.jpg')
```
In a similar fashion you can operate on batch wav files recursively:
```python
>>> import spectrust
>>> with spectrust.Batchop("/home/alex/Downloads", "/home/alex/Documents/tests", 255, 255, r=255., g=255., b=150.) as s:
... res = s
[, , ]
```
The `outputfile` is automatically generated using a UUID4 but it can be changed according to your needs by
extending the `Batchop` class and overriding the generate_output_path method.
### Contribution
Pull requests are welcome but please respect PEP-8 and Numpy docstrings.
### Future development
* Proper documentation.
* Add support for more audio formats and output image formats.
* Tests.
* CI.
* Benchmarking
* Add spectrograms other then mel perhaps...?
Contributions are very much welcome. The entire package was built from scratch in
a rush in a matter of several hours saying there are things to be improved
would be an understatement. Please check the contribution guide.
### Support
As of now the library only supports Linux but it should (in theory)
be working on other UNIX-based systems as well. Running well on
OpenSuSe 15.0 as well as Debian 9.0. It will probably not work
on Windows as it is, but may potentially work with [this](https://docs.microsoft.com/en-us/windows/wsl/about).
*It's a wild guess to say the least...*
Only Python 3.4 and above will be supported. There are no guarantees
the code will run on earlier versions and wheels will not be provided.
### Performace
Not a whole lot of benchmarking has been done for a number of reasons,
time(or the absence of it) being a major contributor.
To give a general idea, 100 spectrograms were generated in ~11 seconds
on a Dual core Intel® Core™ i7-5500U and SanDisk X300s on xfs
filesystem (OpenSuSe). Each of the files was roughly 15 seconds long
and ~300kb.