Deep Learning for Aerial Imagery: DeepForest: Automate tree detection with Pytorch

Have you wanted to get into geospatial deep learning but aren't sure how? This session demonstrates the use of the python library DeepForest for detecting trees (or birds) in high-resolution aerial imagery. DeepForest uses the powerful deep learning framework Pytorch, but reduces the complexity for researchers more focused on forestry and ecology. Using python code in a jupyter notebook, we will demonstrate workflows to run existing tree detection models, as well as how to fine-tune a model for your area of interest. An existed account with Cyverse is necessary to follow along in the jupyter notebook.

This session is part of the NextGen Geospatial workshop series that teaches cutting-edge topics around GIS and remote sensing. Each workshop session is designed to be a discrete lesson where students will walk away with specific knowledge on a tool and resources to explore deeper. Our goal is to demystify vocabulary and show you how to use these tools with straight forward examples. 

We welcome students and professionals from any field that are interested in expanding their geospatial skillset. A variety of skill levels are also welcome, though each lesson will assume the audience has limited experience on the topic. Basic knowledge of scripting languages (mostly python) and some prior geospatial experience will be helpful. Some of the lessons will include gentle live coding, but the focus will be on the big picture of what the code is doing. Jupyter notebooks with pre-written python code will be provided for these lessons.

SERIES: NextGen Geospatial: AI & Cloud tools for Geographic Analysis:

When:  Tuesdays, 2:00 - 3:00 PM, Sep 3 - Oct 29, 2024
Where:  Weaver Science-Engineering Library, Rm 212 and on Zoom 
Instructor: Jeffrey Gillan
YouTube: See below for video links.

Workshop Materials