Skip to content

Automated, scalable, and cost-effective ML on AWS: Detecting invasive Australian tree ferns in Hawaiian forests

Published Article

Hawaii, Global

Publication date: March 29, 2022

File format: URL

View resource

This project showed that the cloud-based ML pipeline could reduce manual labeling work from weeks or months to hours, lowering costs and leading to more timely conservation action.

Subject Tags

  • Data Science and Artificial Intelligence

Summary

The AWS–TNC project used Amazon SageMaker and related AWS tools to automate detection of invasive Australian tree ferns in high-resolution aerial imagery from Hawaiian forests. The workflow divided very large georeferenced images into smaller tiles, trained an object-detection model to locate ferns, and produced coordinates that could help TNC prioritize field treatment more quickly. The blog emphasizes that the cloud-based ML pipeline could reduce manual labeling work from weeks or months to hours, lowering costs and speeding invasive species response.

Citation

by Dan Iancu, Arkajyoti Misra, Theresa Cabrera Menard, Kara Yang, Veronika Megler, and Annalyn Ng 

TNC Authors