Analytic Feeds consist of multiple layers that process information to deliver detailed outputs. The pipeline starts with raw data from satellite imagery, filters out noise to improve quality, and prepares the data for API access. Advanced algorithms then detect and classify objects within the imagery, such as vessels on the sea. The final layer draws bounding boxes around the detected objects to support tracking and analysis.

Vessel Detection is trained to perform reliably across different sea states, atmospheric haze, sun angles, and vessel types. It is most effective for vessels longer than 25 m in open water, and its published accuracy metrics apply only to those conditions. Smaller vessels, and vessels outside open water, are more likely to be missed. Even within those conditions, detection is not exhaustive: in validation, the model identified 84% of the vessels present (recall 0.84), so some vessels are missed even when they look similar to vessels that are detected. For details, see the Vessel Detection technical specification.