This repository provides the source code used to construct Gaussian Process–based stochastic maps derived from image segmentations obtained via CNN-based object detection. At the moment, and during the revission period, it is only available under demand.
This repository provides the source code used to construct Gaussian Process–based stochastic maps derived from image segmentations obtained via CNN-based object detection. Several Files are included: Source code in python and based on Pyro and pytorch libraries for the Gaussian Process, a sample image with their "smoothed" couterparts, and the yml file to generate the conda environment with all packages needed to run it
Take into account that it has been optimized for python 3.8
There are 3 zip. One contains a set of segmented images with inferences given by a CNN. The other two zips correspond to the same images of the same dataset, but with the segmeted infereces smoothed with a regular Gaussian ad with a simetric Gaussian.
Contact us to require a copy of the Python sources and datasets: [email protected]