clean-IT: Towards Sustainable Digital Technologies

Recent developments of mobile-device technology enable the processing of neural networks on-device using specialized hardware, e.g., so-called neural processing units. This allows for the application of neural networks directly on these devices without requiring server-based processing. However, using this hardware often requires adaptation and optimization of neural networks to operate in a power-efficient way. In a first approach, the Visual Media Analysis and Processing Group at HPI’s Chair of Computer Graphics Systems in collaboration with Digital Masterpieces GmbH successfully adapted and optimized convolutional neural networks for image and video analysis and stylization. More information...

Dr. Matthias Trapp is a senior researcher, lecturer, and principle investigator at the HPI's Computergraphics Systems Chair. His research interests includes GPU-based image and video processing techniques and frameworks, interactive GPU-based rendering techniques for massive spatio-temporal data, focus+context visualization techniques for virtual environments, web-based visualization for mobility analytics, and interactive software maps.