Opportunities

We are always happy to receive applications of motivated students for a thesis or a research internship.
If you are interested in doing research in the domains computer science, Machine Learning and Bioinformatics, do not hesitate to contact us. For more information please contact Dominik Grimm.

Opportunities for Bachelor and Master thesis or long-term research internships as well as for research internships can be found after PostDoc and PhD positions.

Bachelor, master thesis or long-term research internships

[Bioinformatics & Image Analysis]: Design of a Web Application for Automatic Seed Germination Assessment

Seed research is mainly done by manually assessing germination experiments, which is time consuming and error prone. Modern machine learning based methods (such as deep learning) have been used in order to automate this process. Currently, this software can only be used by domain experts. The goal of this thesis is to develop a responsive and modern web-application that can be easily used to discriminate the germination status of seeds from images.

Read more

[Bioinformatics & Image Analysis]: Design of a Web Application for Automatic Seed Germination Assessment

Seed research is mainly done by manually assessing germination experiments, which is time consuming and error prone. Modern machine learning based methods (such as deep learning) have been used in order to automate this process. Currently, this software can only be used by domain experts. The goal of this thesis is to develop a responsive and modern web-application that can be easily used to discriminate the germination status of seeds from images.

Read more

[Machine Learning]: Sales and price forecasting for perishable goods by the example of apples

Predicting the future based on historical observations is a common problem in many areas. For this purpose, modern statistical and machine learning based methods for Time Series Forecasting are widely applied. The goal of this thesis is to develop a sales and price forecasting system using classical Time Series Forecasting (e.g. Exponential Smoothing or ARIMA) and Machine Learning (e.g. XGBoost or LSTM) approaches and datasets provided by our partner Agrarmarkt Informations-GmbH. Finally, you should draw a conclusion whether sales and prices of apples are predictable and if the results are likely to generalize for other species.

Read more

 

[Machine Learning]: Application of Time Series Forecasting and Machine Learning 
approaches for horticultural sales predictions

Predicting the future based on historical observations is a common problem in many areas. For this purpose, modern statistical and machine learning based methods for Time Series Forecasting are widely applied. In our research project, we focus on sales of small and medium-sized horticultural companies. The goal of this thesis is to apply already implemented classical Time Series Forecasting (e.g. Exponential Smoothing or ARIMA) and Machine Learning (e.g. XGBoost or LSTM) approaches to datasets provided by a partner company. Finally, you should draw a conclusion whether horticultural sales are predictable and methods generalize well across products and companies.

Read more

 

[Machine Learning]: Evaluation of modern probabilistic Time Series Forecasting approaches for sales of small and medium-sized companies

Predicting the future based on historical observations is a common problem in many areas. Probabilistic methods provide several advantages for this purposes, but modern approaches are usually based on large datasets. These are often not available in small and medium-sized companies. The goal of this thesis is to evaluate whether they are nevertheless applicable for horticultural sales predictions based on datasets provided by partner companies.

Read more

 

[Machine Learning]: Evaluation of global and local pattern approaches for sales predictions of small and medium-sized companies

Predicting the future based on historical observations is a common problem in many areas. A potential way to implement this are Time Series Forecasting methods. Some of these combine global and local patterns present in the data. Global effects might be identified in related time series, e.g. from the same domain. These can be enriched with local patterns, e.g. of a specific company, to provide final predictions. There are several approaches in literature which make use of this idea. The goal of this thesis is to evaluate their applicability to sales predictions of small and medium-sized horticultural companies.

Read more

 

Research internships

[Machine Learning]: Genetic Algorithms – Solve Optimization Problems like Evolution did

This project is about the implementation and application of a Genetic Algorithm. Genetic Algorithms are methods for solving optimization and search problems inspired by biological mechanisms, e.g. used in research for the parametrization of artificial neural networks. You will implement a Genetic Algorithm in Python and apply it to solve problems like breaking a PIN or determine variables of a mathematical equation. An overview of this approach and its components can be found in Lingaraj, H., 2016.

For more information please contact Florian Haselbeck.

[Machine Learning]: Theta Method – Time Series Prediction based on Decomposition

Time Series Prediction has various applications e.g. in predicting sales, energy demand or financial parameters. The Theta Method is a statistical approach, which is using historical observations of a time series and decomposing it based on their local curvatures. Despite its simplicity, the method is still delivering good results in many applications. You will work with Python, evaluate publicly available implementations as well as doing your own ones. For that purpose, you will use public time series datasets. An overview of the Theta Method and an optimization of it can be found in Fioruci, J. et al., 2015.

For more information please contact Florian Haselbeck.

[Machine Learning & Literature Review]: Explainable AI meets Time Series Prediction

Time Series Prediction is a research topic with various applications, but simple statistical methods are still widespread in practical use. One reason for that is the explainability of predictions resulting in trust of the user. Therefore, understanding and explaining Time Series Prediction is an important topic. You will conduct a literature research on existing approaches in the context of explainable AI and Time Series Prediction. The result of your work will be an overview of scientific works in this research area with a short summary of the methodical approach of each paper.

For more information please contact Florian Haselbeck.

[Machine Learning & Image Analysis]: Weed Detection in Precision Agriculture

The goal of this project is to generate a dataset for weed detection using remote sensing. You will work on already existing images in order to segment instances of weeds and plants as well as to implement a semi-automatic process to help accomplish the task. An overview of Instance Segmentation can be found in Hafiz, A. M., & Bhat, G. M. (2020). A survey on instance segmentation: state of the art.

For more information please contact Nikita Genze.

[Machine Learning & Image Analysis]: Seed Germination Assessment

The goal of this project is to extend an already existing dataset for seed germination assessment. You will optimize a setup of seed germination experiments in order to generate high quality captures of the seed germination process. Further, you will apply different image segmentation techniques to generate annotations for those images as well as to generate a pipeline for automatic segmentation. An overview of Instance Segmentation can be found in Hafiz, A. M., & Bhat, G. M. (2020). A survey on instance segmentation: state of the art.

For more information please contact Nikita Genze.

[Machine Learning & Image Analysis]: Seed Germination Prediction using  Deep Learning

The goal of this project is use state-of-the-art machine learning methods (e.g. deep learning) to identify seeds within an image, to predict whether it germinated or not and to predict the seed species.

For more information please contact Nikita Genze.

[Literature Research]: Vegetation Indices – Strength, Weakness and Application

Vegetation Indices can be used as a proxy to evaluate vegetation (i.e. cover and health) and can be used to distinguish between soil and plant. These indices have been widely implemented. The goal of this project is to conduct a literature research on usable indices for detecting plants in images obtained by UAVs (Unmanned Aerial Vehicles). Furthermore, you will compare a selection of vegetation indices and summarize their strengths, weaknesses and applications. An overview of vegetation indices can be found in Xue, J., & Su, B. (2017). Significant remote sensing vegetation indices: A review of developments and applications.

For more information please contact Nikita Genze.

Kontakt

Professorship Bioinformatics

Petersgasse 18
94315 Straubing

Head

Prof. Dr. Dominik Grimm

Phone: +49 (0) 9421 187-230
Fax: +49 (0) 9421 187-285
E-Mail: dominik.grimm@hswt.de

Team Assistant

Anna Fischer

Phone: +49 (0) 9421 187-231
Fax: +49 (0) 9421 187-285
E-Mail: anna.fischer@hswt.de