TensorBank

TensorBank is a collection of assorted algorithms expressed in Tensors.

We do not intend to limit ourselves to a specific domain. The initial batch of algorithms is focused on point and box gemoetry for some object detection tasks, but more algorithms will be added later.

We are open to all backends, including Tensorflow, Pytorch, and NumPy.

Primarily, this project is to avoid copy-pasting the "utils" directory from one project to the next :-)

Installation

$ pip install tensorbank

Usage

If you're using TensorFlow, import TensorBank as follows and use the tb. prefix:

import tensorbank.tf as tb

tb.axis_aligned_boxes.area(
    [[1, 1, 2, 2],
     [-1, -1, 1, 2]])
>>> tf.Tensor([1 6], shape=(2,), dtype=int32)

See API Reference on Readthedocs for the full list of the algorithms offered and comprehensive usage examples.

Development version

If you want to install the push-on-green development version, do this instead:

$ pip install tensorbank-dev