The research center will use the system to quickly analyze large-scale satellite and aerial imagery using image processing and deep neural network training, as well as for various deep learning experiments.
One experiment aims to develop new applications that will support rescuers in disaster-response scenarios by enabling them to make faster decisions. The resulting applications could help answer important questions such as finding the areas affected by a disaster and the accessibility of infrastructure during events such as floods.
Another highly topical research area is to measure and understand convolutional neural networks (CNNs) by quantifying the amount of input the let in. This technology is breaking new ground in the area of neural network understanding, opening a new way to reason, debug and interpret results.
“The analysis of big amounts of data — for example, large-scale aerial and satellite imagery — requires a powerful solution to process and train these deep neural networks,” said Andreas Dengel, head of the research department Smart Data & Knowledge Services at DFKI in Kaiserslautern. “The increased memory footprint of the DGX-2, enabled by the fully connected GPUs based on the NVSwitch architecture, will play a key role for us in improving the development of effective AI applications and expand the unique infrastructure of our Deep Learning Competence Center.”
Founded in 1988, DFKI previously used the NVIDIA DGX-1 for various projects including its DeepEye project. To help estimate and forecast damages of natural disasters, the system trained multiple CNN models to extract relevant information from text, image and metadata from social media.
For more NVIDIA developments at #GTC18, follow @NVIDIAEU.