/show-cases

A page with some examples of the products, prototypes & experiments that we do at KI labs.

Welcome to KI labs showcases


Where Weight

Technical product documentations, like engineering drawings, have standards for sizes and layout of drawing sheets (ISO 5457). Nevertheless, finding and getting information from these unstructured data sources like images, scanned documents is a quite challenging and prevailing issue. Existing solutions like AutoCAD, Adobe or cloud OCR (Optical Character Recognition) don’t work well enough for specific tasks because of their too generic approach. This makes building an OCR (Optical Character Recognition) system for a specific goal relevant and actual task.

With Where Weight we demonstrate an approach of extracting data from various engineering drawings, using different processing frameworks from classical computer vision to modern deep learning tools. As a result our customer

#mercedesbenz #deeplearning #OCR

Learn more about Where Weight ...


What Benz

Automotive giants are already very experienced in marketing their cars from best possible perspectives. Over the years they all developed their ways of giving information to their user base, especially with the help of digital media. But the approach is getting old, since it is focusing on the approach “making people want a car” instead of “making cars people want”.

We twisted that approach for Mercedes Benz Cars to see which car models are the ones people want. Based on that interest we provide users all possible content about that specific model. The user runs our app, makes a picture of a Mercedes Benz and immediately knows the model, its base price, engine options, change the color, and much more. Behind the scenes, a fine-tuned deep-learning model makes it possible to provide you the necessary information in seconds.

#mercedesbenz #deeplearning #imageprocessing

Learn more about What Benz ...


Data Hub

World-wide car manufactures have been building different type of vehicles over 100 years. In the last decades, some of them have been collecting billions of data points from dozens of built-in sensors, from hundreds of thousands cars and time points. They believed knowing where the vehicles are stopping by would open different horizons for their next generation digital services.

Our approach primarily was to get the data, crunch it and detect where car/trucks are idling. As simple it sounds, the challenge was to get clustering right so that it makes sense in terms of geolocation, time of the day and duration.

#anonymization #geolocation #trucks

Learn more about Data Hub ...