At Skogsstyrelsen we are working on several projects using AI methods. Currently we utilize the Microsoft Azure cloud service, but the computationally more demanding projects end up leading to very high running costs on that platform. We would therefore like to explore the possibility of using the excellent national infrastructure here at SNIC for these projects. Further we are interested to transition from a cloud system to a SLURM cluster system.
One of the projects we are currently developing is called "digitala naturvärden". In this project we aim to develop an AI model (convolutional neural network) that utilizes many layers of remote sensing data (e.g. satellite images, airborne laser-scanning, aerial photographs, etc) to predict the natural value (naturvärden) of forests in Sweden. We are training the model with annotated polygons of high-conservation forests (high naturvärden) and production forests (low naturvärden). We are currently implementing the model and are using docker containers for building the environment. We hope to be able to utilize the excellent GPU infrastructure on the Alvis cluster at C3SE for this purpose.
We have been in contact with Marcus Lundberg from NBIS who gave us some advice in which resources may be the most suitable to apply for.
I'm applying as the main PI because I'm currently working closely with Skogsstyrelsen on this project. However, Skogsstyrelsen is interested in exploring the possibility to continue using SNIC resources, also after my affiliation with them is over.