On this page, we offer you an overview of each demonstrator created at ScaDS.AI Dresden/Leipzig. You can check out all of our demonstrators in the ScaDS.AI Living Lab locations.

Our Demonstrators

FloodVis

• Immersive virtual reality experience 
• Based on existing simulation data and geospatial information
• Preprocessing of data using elevation data, areal images, water level, and building information 
• Applying areal images as texture on terrain layers and water surface for increased visualization accuracy
• Interactive environment with intuitive navigation methods 
• Enhanced navigation mechanism to avoid motion sickness

Creators: Marzan Tasnim Oyshi, Verena Maleska

Find out more about FloodVis

asanAI

• No-code toolkit to design, train and test machine learning models
• No programming knowledge required
• Runs in browser / no installation required
• Works entirely offline
• User privacy is guaranteed
• Single-click automatic generation of Python/NodeJS code
• Suitable for beginners as well as expert users

Creator: Norman Koch

Find out more about asanAI

Try it yourself

MultiCut

• Gamification of multicut problem 
• Providing an intuitive and interactive way to understand a data segmentation task
• Score-based approach to guide the players optimizing their selected multicut 
• Using both scientific and non-scientific examples 
• Developed in Python, runs on both Linux and Windows

Creator: Jannik Irmai

Big Data Performance Analysis

• Big data framework internals are often hidden form the user 
• Monitoring internal behaviors such as memory access patterns are crucial for analyzing the application’s performance 
• Measurement codes are added to Apache Spark and Flink
• Both tracing and profiling methods are supported 
• Providing side-by-side comparison of performance metrics
• Functions of interest can be identified by the user

Creator: Jan Frenzel

OmniOpt

• Automatic hyper-parameter optimization for a wide range of problems
• Handling of use-case as black box 
• Highly parallel approach (available on Taurus HPC) 
• Stochastic optimization algorithms (TPE, Hyperopt) 
• Capable of handling a large number of parameters
• Automatic evaluation and plotting functions

Find out more about OmniOpt

Creators: Dr. Peter Winkler, Norman Koch

Napari-KICS

• Estimating chromosome sizes using image processing 
• Applicable to all karyotype images
• Automatic identification and labeling of chromosomes
• Intuitive graphical user interface to enable manual adjustments 
• Providing multilayer representation of chromosome characteristics 
• Freely available as a plugin for Napari

Creator: Arne Ludwig

Content Based Image Retrieval (CBIR)

• Retrieving images similar to an input image 
• Based on the on-the-fly calculated similarity scores 
• Machine learning model trained on a large set of historical images from the city of Dresden 
• Fast identification with low processing overhead

Creators: Dr. Christoph Lehmann and Dr. Taras Lazariv

Bridging between Data Science and Performance Analysis

• Bringing performance analysis methods to the world of data science 
• Providing precompiled jupyter kernels for performance analysis 
• Providing Score-P Python binding to analyze performance of Python codes in Jupyter notebooks 
• Enabling global as well as cell-based tracing of Python code performance
• Recorded traces available for Vampir (OTF2)

Creators: Elias Werner, Lalith Manjunath, Jan Frenzel, Dr. Sunna Torge

ImageSeg

• Applicable for all image segmentation tasks
• Enhanced annotation mechanism to improve task automation
• Intuitive and no-code graphical user interface 
• Batch processing of unlimited number of images 
• Single-click execution in browser (no installation) 
• Short return time, thanks to its execution on Taurus HPC cluster

Creators: Dmitrij Schlesinger, Norman Koch, Dr. Peter Winkler, Patric Röhm

CodeCAI

• Dialogue system for creating an analysis workflow
• Reducing learning overhead 
• Simplifying handling of libraries 
• Explaining analysis results in an understandable way 
• Automatic mapping of natural language to Python code

Creators: Bernhard Stadler and Klaudia Thellmann

Find out more about our research!