FastGeodis: Fast Generalised Geodesic Distance Transform

The FastGeodis package provides an efficient implementation for computing Geodesic and Euclidean distance transforms (or a mixture of both), targeting utilisation CPU GPU hardware. In particular, it implements the paralellisable raster scan method from Criminisi et al. (2009), where elements in row (2D) or plane (3D) can be computed with parallel threads. This is able to handle 2D as well 3D data, achieves up 20x speedup on 74x compared existing open-source library (Wang, 2020) that uses non-parallelisable single-thread implementation. performance speedups reported here were evaluated using volume data Nvidia GeForce Titan X (12 GB) 6-Core Intel Xeon E5-1650 CPU. Further in-depth comparison improvements are discussed documentation: https://fastgeodis.readthedocs.io