Ankit Kariryaa: Deep Learning for digital twins of individual trees

Recent advances in remote sensing and machine learning have made it possible to create digital twins of individual trees, providing unprecedented insights into their health and dynamics. My presentation will focus on the application of deep learning techniques in tree mapping and forest segmentation. 

I will showcase how neural networks are used to segment tree crowns from satellite imagery, enabling precise measurements of canopy area and tree distribution within and outside forests. Additionally, allometric equations will be applied to estimate biomass and stored carbon levels for individual trees.I will also explore the critical aspects of large-scale image processing pipelines and the selection of optimal satellite images for accurate mapping of individual trees across vast areas. 

Finally, I will delve into the self-supervised learning methods, which enable models to learn from large amounts of unlabeled data. By harnessing the vast quantities of satellite imagery and other remote sensing data, we can train models that can recognize patterns and features relevant to tree monitoring without human annotation.

Ankit Kariryaa is a tenure-track assistant professor of bridging geo- and AI-sciences, embedded in both the Department of Geosciences and Natural Resource Management and the Department of Computer Science at the University of Copenhagen. Ankit received his Ph.D. in Computer Science from the University of Bremen, Germany in 2020 with highest distinction. He is deeply passionate about advancing machine learning techniques to address environmental challenges and combat the climate crisis