Masters Theses
Are you a master's student eager to delve into cutting-edge research and make a meaningful contribution to emerging fields remote sensing with computer vision? Our chair invites you to write your master's thesis with us, offering a dynamic and supportive environment where your ideas can thrive.
What we offer
We offer you the opportunity to use the fast resources of our chair not only to complete the last step of your Master's degree, but also to contribute to an existing field of research and possibly start your way to a PhD position with a high-ranking publication. Specifically, we offer supervision by a multidisciplinary team of data scientists and economists, the opportunity to integrate your research into a larger research project, and powerful computational resources. We have extensive in-house datasets and deep learning models that we are happy to share. Students should be trained in remote sensing, econometrics, GIS, data science or similar. Your specific academic background is less important than research interests that align with ours, a high level of motivation and the technical skills you bring to the table. If you are interested, please fill in the form at the bottom of the page.
Current opportunities
To get a glimpse of our current projects have a look polybox (ethz.ch), external page [2303.02230] Building Floorspace in China: A Dataset and Learning Pipeline (arxiv.org) and external page Urban Growth Unveiled.
We are currently looking for students to take on the following challenges:
- Use a XGBoost-model for post-processing of building height prediction similar to external page essd-2024-217.pdf
- Temporal smoothing for post-processing of building height prediction (in principle we are open for any unsupervised method)
- Enhancement of the current height and footprint prediction model (CNN) on optical satellite imagery using batch input
- Extension of a current deep learning model that extracts road networks from road maps in China. We have two possible extensions in mind: a. adding satellite imagery, b. creating synthetic training data with different map symbologies.
- Implement a deep learning model (CNN) that predicts construction volumes based on Landsat 5 satellite imagery.
Needed skills and attributes:
- Very good Python (or R) skills and ability to write efficient code for scalability
- Prior knowledge of computer vision
- Highly motivated to dive into unknown territory with uncertain outcomes
- Prior knowledge of GIS is a plus
- Willingness to turn the thesis into a publication is a plus
How to Apply
If you are interested in joining our group for your master's thesis, please fill in the following form: external page https://forms.gle/JVG4w1KBnjRsKX8X6. Non-ETH-students are welcome to apply, please make sure that your study program agrees to external master-thesis supervisors. If there are any questions, please contact Sebastiano Papini: