Deutsche GigaNetz GmbH

Project result: Poster (PDF)

Training of an AI for surface analysis

Description of the companyDeutsche GigaNetz GmbH was founded in 2020 and has set itself the goal of building the next-generation fiber-optic network.

The young company is growing rapidly, now has around 500 employees across Germany and aims to grow to 1,000 in the next few years. It is one of the fastest growing companies in Hamburg.

In a highly competitive environment, Deutsche GigaNetz GmbH has asserted itself in a short time and offers high-speed Internet with giga bandwidths for companies and private households throughout Germany.

Situation
For the private-sector expansion of the fiber optic network, identifying and evaluating potential build-out areas is an essential part of the business model. Millions of data records are collected, analyzed and visualized for this purpose. Potential build-out areas are transferred to a portfolio from which the sales department can derive the next expansion projects in each case.

A key component of this process is surface analysis, for which new approaches are being sought as part of this project.

Problem
The identification and selection of suitable areas as well as the cost-benefit analysis for the build-out are associated with great effort. There is a lack of data sets that represent the roads and sidewalks as polygons. Roads and sidewalks are mainly available as lines, e.g. OSM, and are partly incomplete.

The existing datasets do not contain any or insufficient information on the surface covering of the roads and sidewalks. Corresponding data on roads and sidewalks, including surface coverings, must currently be purchased at high cost from external companies.

Aims of the project
The project aims to find new and creative approaches to classify surfaces and other objects. The goal is to develop a suitable method to combine existing digital resources and train an AI system accordingly.

Primarily, the segmentation of roads and sidewalks as well as the classification of the surface into the categories asphalt, pavement or unsurfaced is concerned. In addition, the detection of trees represents a significant aspect of surface analysis. So far, the latter has not been feasible with reasonable effort.

On the part of Deutsche GigaNetz GmbH, the following digital resources are available for a concrete build-out project:

• Driving data (laser data)
• Images on site
• Orthophotos with 20 cm resolution

Scopes

• Strong Affinity to data science
• Computer vision
• Machine Learning
• Experience with geo data
• Image processing
• Pattern recognition
• Programming skills (no specific language)

Target group (students)
Data Science, Computer Science, Informatik-Ingenieurwesen, Bau- und Umweltingenieurwesen, Elektrotechnik, Technomathematik, B.Sc. and M.Sc.

Dates
Please save these dates: Fishing for Experience Termine

Registration
You can apply for Fishing for Experience online.