We deal with topics around Big Data and AI, because we believe that only large and well-researched data are a basis for modern AI-driven analysis and enable new insights that were not possible before. Just follow our new team member, the Coding Cat, on a short tour.
This is my friend Schroeder. He is quite nice, although he is only a dog. Paul is his master and works with lots of data and neural networks. Sometimes Paul has no time for Schroeder, as he sits on his computer all day. But I think we at ScaDS.AI can definitely help him! We have a solution for the training of neural networks and hyper parameters. Find out about it by clicking on the video! And here for more information …
We also work with image segmentation. In our case, segmentation means the division of the image into content-related regions, whereby the criteria are defined by the user. Examples are the distinction between foreground and background, between certain characteristic areas in topographic maps or satellite images or between “interesting” and “uninteresting” areas in microscope images. Automated digital image segmentation is usually performed using machine learning (ML) methods. The demonstrator for image segmentation allows users to segment their own images according to criteria they specify. Have a look here!
Our image retrieval pipeline is implemented using deep convolutional neural networks and it takes solely pixel information of an image and does not rely on any available metadata. The basic idea is, that we have the query image with the object of interest and many reference images from an archive to compare with. The “distance” between the query image and every reference image is calculated which allows to have a ranked list of most similar images and to find the closest match. A demonstrator for image retrieval was developed at the ZIH and is available. Please contact us for more information!
We are assessing future flood risks using climate change as driver for the entire flood risk chain comprising atmospheric, hydrological, hydrodynamic and damage processes. By multi-model applications the climate change signal is routed through every component of the risk chain and uncertainties stemming from climate scenarios, climate models and hydrological parametrisations are identified and quantified. Different climate change ensembles serve as input for long-term high-resolution hydrological modelling to detect future flood events. Inundation maps are derived by 2dimensional hydrodynamic modelling.
Damage estimation covers construction and inventory damage. Transient risk curves including uncertainty bounds are derived. Due to the high-resolution in space and time high performance computing is applied. Performance testing is crucial to run a comprehensive model chain. A special challenge is the coupling of the model chain, since different system requirements have to be coordinated. Visual data analytics support to understand responsible processes of flood risk generation in a flood risk system. Some images do not even have metadata, because the photographer, object, period, or other key data may be unknown. Here you can find more information.
Just send us a message! 
We appreciate it!
OpenMoji License: CC BY-SA 4.0