Computer Vision for Geospatial Data

About this course

Apply computer vision concepts to aerial and drone imagery, from tiling and labelling to segmentation, vector outputs and accuracy assessment.

This advanced course is part of the GeoAI & Geospatial Intelligence school. Work through the curriculum below in a instructor-led / cohort format, with an estimated study time of 23 h.

Course curriculum

• Module 1 — Aerial and Drone Imagery
• Module 2 — Tiling and Labelling Strategies
• Module 3 — Semantic Segmentation
• Module 4 — Object Detection
• Module 5 — From Pixels to Vectors
• Module 6 — Object- and Area-Level Accuracy
• Module 7 — Spatial and Domain Shift
• Module 8 — Deployment and Inference Cost

Applied learning

• Case Study — Pennard: A Solar Register That Worked in June
• Lab 1 — Segment buildings and vectorise them
• Lab 2 — Detect solar panels and measure object-level precision and recall
• Lab 3 — Test a model on a town it has not seen
• Capstone — A Geospatial Vision Product You Can Defend

Who this course is for

GIS / Remote Sensing Professional; Data Professional; Researcher

Prerequisites

NE-AI-106

Course details

• Course code: NE-GEO-305
• Level: Advanced
• Estimated duration: 23 h
• Delivery: Instructor-Led / Cohort