Home

This isetcloud toolbox simplifies running docker images managed by kubernetes (k8s) from Matlab. This implementation is designed to for the Google Cloud Engine on the Google Cloud Platform (GCP), which requires having the Gcloud SDK installed on your system. See the installation instructions. The isetcloud software fits within the general iset compute environment like this:

We initially developed isetcloud for a computer graphics rendering application (see iset3d). In that application, we implemented PBRT in a docker image so that PBRT could be run using the kubernetes system. This enables us to increase the number of compute nodes when we want to render many different versions of an image.

Recently, we extended isetcloud to help us run certain machine learning applications. In this case, we move data to the GCP to train and evaluate TensorFlow models. (Henryk Blasinski, Zhenyi Liu).

The isetcloud toolbox may have additional use cases. For example, we expect to use this toolbox for neuroimaging and vision science simulation applications from its origins in rendering.

Programming note: The namespace for isetcloud programming is defined by prepending 'mc' to most of the functions or 's_mc' to the scripts. The reason for this choice is we initially called isetcloud by the name matlab2cloud, and thus 'mc' made sense. When we renamed the repository isetcloud, which is less general and more appropriate to what we are doing, we left the namespace as 'mc'.