chunky3d 0.1.13

A 3D array-like NumPy-based data structure for large sparsely-populated volumes

chunky3d

A 3D array-like NumPy-based data structure for large sparsely-populated volumes

Build

Introduction

This library provides a data structure, Sparse, which represents 3D volumetric data and supports a subset of np.ndarray features.

Example

>>> import numpy as np
>>> from chunky3d import Sparse

>>> s = Sparse(shape=(64, 64, 64))
>>> s[0, 0, 0]
0

>>> s.dtype
numpy.float64

>>> s.nchunks
8

>>> s.nchunks_initialized
0

>>> s[1, 2, 3] = 3
>>> s.nchunks_initialized
1

>>> s[:2, 2, 3:5]
array([[0., 0.],
       [3., 0.]])

Features

• chunky3d.sparse_func - a collection of functions for analyzing chunked arrays, including morphological operations (opening, closing), thinning, connected components
• Fast load and save using msgpack
• Operations on arrays using .run(), with possible acceleration using multiprocessing
• multiprocessing-based acceleration in most of existing sparse_func
• Accelerated lookup using numba
• Interpolation (point probe)
• Origin and spacing: representing 3D space with non-uniform spacing for different axes
• Easy visualization of arrays with dtype=np.uint8 via chunky3d.k3d_connector.get_k3d_object()