hi, i'm linda and i create things.
hey there!

I'm Linda (26 years old) and currently studying Data Science at the ZHAW School of Engineering in Zurich, Switzerland.

I am passionate about programming, data science, computer vision and making music.

You can check out my LinkedIn, my Github or get in touch with me by email. Let's talk!

get in touch

Volumetric video capture
Master thesis

Volumetric video defines a technique that captures a three-dimensional space, which later can be viewed in 2D (or 3D with VR goggles). The focus in my master thesis is on a 2D rendering output, however the environment as well as the viewpoint of the camera can be changed interactively by the user. In the thesis I concentrate on three different aspects:

  • 1. Setup the capturing environment with up to three RGBD cameras using the Imverse API.
  • 2. Implement a virtual camera in Unity to change the viewpoint of the hologram
  • 3. Improve the quality of the captured object
Learning to Ignore: Fair and Task Independent Representations
Project work

Training fair machine learning models, aiming for their interpretability and solving the problem of domain shift has gained a lot of interest in the last years. There is a vast amount of work addressing this topics, mostly in separation. In this work we show that they can be seen as a common framework of learning invariant representations. The representations should allow to predict the target while at the same time being invariant to sensitive attributes which split the dataset into subgroups. The approach is based on the simple observation that it is impossible for any learning algorithm to differentiate samples if they have the same feature representation. This is formulated as an additional loss (regularizer) enforcing a common feature representation across subgroups. It is applied to learn fair models and interpret the influence of the sensitive attribute. Furthermore it can be used for domain adaptation, transferring knowledge and learning effectively from very few examples.

Read paper
Dynamic division of Switzerland
Project work

I did my first project work with the title "Dynamic division of Switzerland into areas with homogeneous weather conditions based on air pollutant concentrations" during my masters in the statistics field. The pollutants in the air, which are mainly produced by humans, have negative effects on humans and the environment. In order to protect risk groups and the sensitive ecosystem, and to monitor exceedances of pollutant concentrations, a large network of air monitoring stations is necessary. The operation of these monitoring stations can be expensive and failures and malfunctions of the stations can occur again and again. Furthermore, the measurements are only carried out selectively and not for entire regions. The aim of the work is to provide a comprehensive statement on air quality for the whole of Switzerland and to investigate certain conspicuous locations more closely.

Webshop for sports equipment

A webshop being developed for aeddict.

With its online store concept in the sports sector aeddict addresses the needs of its customers with the help of selected partners and evolves the shopping experience on the internet. A small questionnaire helps the customers find the suitable products for their sports hobby.

The website is partly developed in Java and Vaadin as well as in Shopify's "Liquid".

Secure authentication of drones
Bachelor thesis

It is expected that within the next several years the number of drones buzzing around will increase steadily. To ensure safety and efficiency of air traffic, drones need to be reliably identifiable and localizable as well as protected against deliberate manipulation. The currently used unsecured data protocol allows different attacks on the system that might even cause dangerous situations. This thesis gives an answer to the following question: Is it possible to design a secure public-key infrastructure for drones? A prototype of this system was implemented with the support of FLARM Technology AG in the form of a web application. It is possible to verify drone packets which use the "FLARM UAS Electronic ID" protocol, via a REST interface, register drones and their opera- tors using public-key procedures (EdDSA) and imitate user authentication via SMS. Additionally the current validation states of the drones as well as their positions are visualized during flight in Google Maps. It is demonstrated how this process can be integrated into the infrastructure so that the chain-of-trust works. To design the prototype, different infrastructure models were analyzed and the best approach chosen. All entities involved in the system were evaluated and their architecture along with the main processes such as registering a drone or verifying drone packets, determined. Lastly, the apparent vulnerabilities of the chosen infrastructure model and the resulting possible attacks were identified.

Image processing and watches
Accuracy measurement of mechanical watches via image processing

Mechanical watches must be checked for rate deviations both during manufacture and use. Most of the equipment used for this purpose is expensive. Almost all of the devices on the market apply acoustic measurements and a few complement this measurement with an optical measurement done with a laser beam. In this project the question has been tackled whether a Raspberry Pi and a Raspberry Pi camera can be turned into a cheaper and simple measuring device that can compete with the accuracy of a professional device using a new approach: image processing.

Retro browser game

Geroids is an arcade game that follows the retro trend. The design is minimalistic and in retro style. Emphasis is placed on intuitive game operation, the user should not have to read a long instruction manual. In Geroids, the player controls a rocket in space with which he must evade and/or destroy geometric figures. The game was developed as part of a project during studies.

App for childcare administration

The main goal of kits is the simplification of the complicated communication between child care centers and parents. Kits simplifies the data collection in one application, which is running on computers and mobile devic-es. Parents can easily enter the absences and ask for additional time slots to bring in their children. The application was developed as part of a project during studies.