Research Computing Center

Introduction to Deep Learning Models in GIS

Deep learning models detect features in images by using deep neural networks based on training datasets. Unlike traditional imagery models in GIS, deep learning models don't just look at individual pixels or groups of pixels but have a greater capacity for learning and can recognize complex shapes, patterns, and textures within images at multiple scales. As a result, deep learning models can learn from vast amounts of training data under varying conditions and could be reused for prediction to a wide variety of imageries at a much lower computational cost.

In this workshop, we’ll learn how deep learning works through hands-on exercises. We will use a Jupyter notebook and the arcgis.learn module. The arcgis.learn module includes over fifteen deep learning models that support advanced GIS and remote sensing workflows. We will go over object classification, pixel classification, and object detection models. We will also import any prediction or classification model from the popular scikit-learn library for use in the workflow.

Objectives:

• Understand the fundamentals of deep learning and its application in image analysis for GIS.

• Gain hands-on experience using Jupyter notebooks with the arcgis.learn module.

• Explore various deep learning models for object classification, pixel classification, and object detection.

• Learn to integrate prediction and classification models from the scikit-learn library into GIS workflows.

• Recognize the advantages of deep learning models over traditional imagery models in detecting complex patterns and textures.

Level: Introductory

Duration: 2 hours

Prerequisites: Basic understanding of GIS, and python Programming. Participants are encouraged to bring their laptops for an engaging and immersive learning experience.