FloorplanAnalytics 0.0.4

A library to analyze floor plans with common space syntax metrics.

Floor Plan Analytics Package

Tested Python versions are 3.9, 3.10, 3.12 and 3.13

Description

This is package that includes some of the most used Space Syntax tools to be applied on floor plans or floor plates. The package is build to utilize Numba to accelerate calculations of the analytics either by CPU multithreading or by utilizing JIT compiled CUDA kernels. The Numba kernels are exclusively written for CUDA, therefore a Nvidia GPU and CUDA installation are needed for running the GPU accelerated version of the code.

The package can also be installed form PyPI:

pip install FloorplanAnalytics

IIf you receive warnings that Numba can not recognize CUDA and falls back and the GPU methods or CUDA is recognized but the kernels terminate without a warning, try to install the correct numba-cuda version according to your install CUDA version. For updating to the correct numba-cuda after the package installation is finished just run this to overwrite the generic numba-cuda installation (here with CUDA 13 as an example)

pip install numba-cuda[cu13]

0.0.3 Features

Features for the new version are:

• Revisited all GPU methods for speed gains and/or reducing memory consumption
• Added a new method for calculating shortest distance to an obstacle only if a ray hits it

Details

The inputs and outputs of the outputs fo the methods are all handling simple to manage Numpy arrays. This choice is due to two main considerations aimed to make the backend functions as broadly applicable as possible.

1. The Input can therefore stem from a (flat and regular) point grid or directly from an image.
2. The output can be used for further data driven analytics, comparing or storage without any additional data structure

To maintain close-to-realtime performance also at larger array sizes, the exposed methods for the analytics calculations include downscale and up sample functionality while keeping the returned values also scaled accordingly.