Anomaly Watch

This app shows anomalies and change detection over satellite imagery, using two kinds of detection. Both run on embeddings computed on CloudFerro infrastructure, and the written case descriptions are produced by an AI agent.

Patch detection is a Near Real Time simulation: it flags individual scenes as soon as new satellite data comes in. It catches things like wildfires, floods, new construction and cleared forest. Each highlighted patch is about 2.2 × 2.2 km on the ground. It runs on TerraMind-Large-S2L2A embeddings, computed on CloudFerro infrastructure.

Pixel detection compares imagery for the same spot across different periods, usually years but sometimes shorter, to show how the whole area has changed over time. It runs on Tessera v1.1 embeddings, also computed on CloudFerro infrastructure.