New AI system called SkyScraper uses multi-agent feedback to detect news events in satellite imagery 5x more effectively than traditional methods

SkyScraper is an iterative multi-agent workflow that geocodes news articles and generates captions for satellite image sequences to identify real-world events

The system addresses the lack of multi-temporal event captioning datasets in remote sensing by automating the labor-intensive process of searching for visible events and labeling sequences

SkyScraper successfully identifies 5x more events than traditional geocoding approaches, proving that agentic feedback is an effective strategy for detecting multi-temporal events

Researchers created a new multi-temporal captioning dataset containing 5,000 sequences by applying the framework to a large database of global news articles