Case study · NOAA / NESDIS (EO-DT)

The NOAA satellite
anomaly-detection case study.

A ResNet-18 model that flags scene-level anomalies in NOAA GOES and JPSS satellite imagery — 99.2% accurate on GOES ABI, running in production on GOES-16. Tell us where to send it and we'll take you straight to the write-up.

WHAT YOU'LL GET

Read it now,
and in your inbox.

Enter your details and we'll open the case study and email you a copy, straight from David Daniel, our CTO — reply and it comes right back to him.

Inside the write-up

  • › The problem: scene-level defects that pixel QC misses
  • › Why a CNN (ResNet-18), and how it was trained
  • › Grad-CAM explainability for an auditable model
  • › Deployed to GOES-16 production on AWS

No spam. We follow up only about work like this.