In September 2022 I met a group of PhD students from UC Berkeley, California at Machine learning workshop in Glaciology at Finse. After a week, we decided to start collaboration and work on different projects. However, possibility to be able to sit together and efficiently work side by side seemed to be as a best solution for increasing our efficiency. Therefore, I applied for travel grant and spent 3 weeks at UC Berkeley at the Statistic and Astronomy departments where I had a chance to meet many more students and professors for increasing collaboration between institutes and countries.
I am working on my routines for proposing methods for creating more efficient and high-quality final products delivery withing geoscience. The main aim is to produce model that is ideally partially reusable for different purposes within geoscience, geology. I got a chance to join the group of Fernando Perez at the Department of Statistics and work closely mostly with Ellie Abrahams from Department of Astronomy. Having the opportunity being there in person makes discussion and relationships closer also for future collaboration withing institutions. I also got a chance to present my work for the departments and opened for very good discussions.
Before this opportunity I was working on cloud-based solution using Google Collaboratory and Google Earth Engine using python and ArcGIS Pro and Jupyter Notebook. The cloud-based platform is very useful for the calculation and data availability. However, we approached pour new project using Jupyter hub with the support from Amazon Research grant using Tensor flow and PyChram and with the use of Amazon’s powerfull computational machines. The main goal for research stay was working on design, data preparation, statistics behind deep learning and U-net and understanding every single step needed in this process its implementation for geological use.
Within the scope of 3 weeks situated at Berkeley, I can confirm the actual need of in person contacts in order to create contacts for future collaboration and successful projects from start to the end and efficient workflow. The exchanged contributed to the interdisciplinarity of my research, exposed new perspectives to my project same as increasing collaboration between institution in the future. After 2 very efficient weeks we decided to continue with the collaboration and we applied for Peter Sether Grant (https://sathercenter.berkeley.edu/peder-sather-grant/) from UC Berkeley that can affiliate next 2 years of our international collaboration between institutes. Also, we started to write a paper about needs of application of deep learning techniques in a real world with a real (not just a perfect) data, same as a paper with working title: “Automatic delineation of geological lineaments using Deep Learning techniques”.
Machine and deep learning are a very sufficient ways of creating geospatial products considering both map production efficiency and the quality of the product in the future. Machine learning with the help of experts can become a future in producing beneficial products for having better overview about sustainable resources and potential natural hazards.
This is just a beginning, and we are working further with our project. The methods implemented during the exchange stay will be used for further research. In person discussion with experts with different background at the Department of Astronomy and Statistics, University of California at Berkeley contributed to interdisciplinarity of my research, expose new perspectives to my project same as increasing collaboration between institution in the future.
