Haslerstiftung: LinMP: Massively Parallel GPU Parallelization
A LINQ-style programming framework for the .NET framework that transparently enables massive parallelization on the GPU. Project duration 2018-2019.
CTI Project: Optical Packaging Quality Control with Deep Learning
A new solution for automatic optical packaging control based on deep learning, together with and to be applied to industrial partner Intigena Hygienic Solutions.
CTI Swiss Commission of Technology and Innovation. Project duration 2017-2019.
EU Eurostars Project: DeepSegment: Segmentation in Imaging and Biological Imaging.
This is a joint project with Swiss and Swedisch universities and companies. Our lab focuses on large-scaled parallelization in this research project.
Funded by EU Eurostars Program, Consortium of Swiss and Swedish Research and Industry Partners. Project duration 2017-2020.
EU Eurostars Project: XamFlow: A Workflow-Based Examination System for the Complete Micro-CT Evaluation Process
A new workflow-based examination system for the complete micro computer tomography (CT) evaluation process, involving massive-scale medical image processing. The HSR Concurrency Lab implemented the high-performance parallelization of massive image processing on heterogeneous clusters.
Funded by EU Eurostars Program, Consortium of Swiss and Swedish Research and Industry Partners. Project duration 2015-2017.
CTI Project: A .NET Extension for Radically Simplified GPU Parallelization in .NET on the Basis of Alea.cuBase
A new programming model and runtime system atop the .NET framework to make GPU parallelization as simple as possible for a broad mass of .NET developers, without limiting generality and high performance.
Funded by CTI Swiss Commission of Technology and Innovation. Project duration 2013-2016. Project page
Parallelization of Medical Imaging on Clusters
In collaboration with Lucid Concepts AG, a Swiss medical software startup company, a distribution framework has been realized in .NET for processing fine-resolution CT images on MS HPC clusters. The system computes optimal parallelization of graph-like pipelines of processing tasks on the cluster by reducing network traffic of the large CT data as much as possible.
Funded by CTI Swiss Commission of Technology and Innovation. Realization in 2013. Project description
Optimizing for GPU Architectures
For Quantalea, the creator of the F# GPU programming library Alea cuBase, we performed a study on how GPU programs written in .NET could be efficiently mapped to the GPUs. A project continuation is planned.
Funded by CTI Swiss Commission of Technology and Innovation. Realization in 2013.