I recently had the opportunity to intern at ETH Zurich’s Studio Davos/SLF, working on a project that blends geology with machine learning. My background is in environmental sciences, and during my master’s, I developed an interest in applying machine learning to solve real-world environmental problems. While I had worked on projects such as predicting the biodegradability of chemicals, computer vision was relatively new to me. This internship allowed me to explore it further while tackling a critical problem: identifying fractures in borehole imagery.

The project I worked on was focused on advanced fracture identification in borehole imagery through semantic segmentation. While this may sound technical, its significance spans several industries. Boreholes are drilled into the Earth for various purposes, and understanding the fractures and changes in the surrounding rock mass is crucial for making informed decisions about construction safety—whether for tunnels, dams, or even storage facilities for hazardous materials. Understanding how fractures form and evolve, especially in response to factors like ventilation or pressure, can significantly improve safety and efficiency in such projects.

The main challenge of the project was the need for more labelled data. Although there are plenty of borehole images, there is a scarcity of labels that help us differentiate between different types of fractures. Therefore, my initial task involved learning how to identify and label these features accurately. Although labelling borehole images might not seem thrilling, it is critical for improving the model. Once we expanded the dataset and refined certain structural aspects of the model, we observed exciting advancements in its predictive capabilities.

One of the highlights of my internship was the people I had the chance to work with. The team at SLF was made up of researchers and professionals from all over the world, each bringing a unique perspective to their work. Whether it was in geology, environmental science, or avalanche monitoring, the diversity of expertise made every conversation an opportunity to learn something new. Working in Davos also enhanced my overall experience. Whether swimming in the lake during lunch breaks or going on after-work hikes or bike rides in the mountains, the combination of fresh air, exercise, and stunning views made Davos an incredible place to live and work.

Overall, my internship at SLF not only gave me hands-on experience in computer vision but also broadened my understanding of how machine learning can make a real-world impact, even in fields I hadn’t considered before. Moving forward, I’m excited and motivated to continue exploring the intersection of environmental science and machine learning.

Interested in ETH Studios? Discover more here.