tfops Documentation

What is tfops?

The tfops package is a free Open Source collection of Tensorflow ops in Python. They are mostly related to tasks in signal and image processing and primarily intended for research purposes and not for high-performance Tensorflow model services used in applications.

The main features of tfops are:

  • 1D and 2D Forward and Inverse Discrete Wavelet Transforms
  • Non-linear thresholding utility functions (e.g. for sparse regularization)


The documentation of tfops is organized into sections:


Currently tfops supports Python >=3.4 (Python 2 support might be added soon). It is dependant on Numpy, Tensorflow, and PyWavelets.

Tested with these versions (as of July 2018):

Install the required packages (using pip, conda, or any other way you prefer). Then get the tfops package from this repository. The easiest way is to clone the repository into your Python package directory. At the moment this still has to be done by hand (sorry…). Install scripts will be provided soon ;)

Development & Contributions

I started tfops in 2018 as a collection of convenient Tensorflow ops frequently needed for my academic research. As such it is primarily intended for my personal use. Use at your own risk.

Contributions including bug reports, bug fixes, ideas for new features, as well as their implementations and documentation improvements are welcome.


Use the GitHub Issues to post your comments or questions.


The tfops project code is open source and available on Github under the MIT License.