Joe Morrison, Author at Azavea

Slider for testing seams of labeled machine learning data.

3 Ways to Analyze the Results of a Supervised Machine Learning Model

By Joe Morrison on December 13th, 2019

How accurate are our supervised machine learning models and what are they really doing? We offer 3 tips to help you better understand these models.

Software Engineering

Betting Big on the Spatiotemporal Asset Catalog (STAC) Standard

By Joe Morrison on November 22nd, 2019

We're investing heavily in the STAC specification - including building a STAC-compatible Python library and server as well contributing to the Label Extension. We're hoping this work will help accelerate adoption across the geospatial engineering community more broadly.

Screenshot of labeled satellite imagery in Raster Vision.

Company

An Introduction to Satellite Imagery and Machine Learning

By Joe Morrison on November 5th, 2019

Today, the availability of satellite imagery still far outpaces our capacity to analyze it, but machine learning and tools like Raster Vision are helping.

Satellite imagery displaying machine learning labels.

Software Engineering

Azavea and CloudFactory: Partners on Quality Training Data and Social Impact

By Joe Morrison on October 8th, 2019

Read about our journey to hiring an outsourced data labeling firm and how we’ve found a partner in CloudFactory.

Picture of dice overlaid with charts from Anscombe's Quartet.

Software Engineering

An Introduction to Machine Learning Accuracy Metrics

By Joe Morrison on July 31st, 2019

How accurate of a machine learning model can you make? It depends on how we decide to define "accurate".

Screenshot of a machine learning model’s incorrect labeling of a rooftop parking lot

Software Engineering

Accuracy is in the Eye of the Beholder

By Joe Morrison on February 13th, 2019

Accuracy in deep learning models is not as cut and dry as many present it to be. We examine several examples where accuracy is more of a judgment call.