Steller Sea Lion Artificial Intelligence Image Processing and Counting Enhancement Effort

The Performance Work Statement (PWS) from NOAA Fisheries outlines a project aimed at enhancing the monitoring of protected Steller sea lion populations in Alaska using artificial intelligence (AI). The Alaska Ecosystems Program seeks to refine its aerial imagery collection processes by improving current image registration and detection models employed for counting sea lions. Existing methods are time-consuming and somewhat inaccurate, involving traditional manual counting that can take up to four months. The project will focus on further developing two key models: the FUSION detection and classification model, which aims to improve accuracy in identifying individual sea lions, and the Suppression image registration model, meant to address overlapping imagery issues. Additionally, the contractor will refine relevant pipelines, scripts, and the DIVE GUI to enhance user interaction and data processing. Key deliverables include quarterly virtual meetings, monthly performance reports, improved tile mosaic capabilities for image sets, model assessments, and virtual training sessions. The overall goal is to streamline the monitoring process, facilitate data analysis, and ensure effective conservation efforts for the Steller sea lion populations while monitoring performance through stringent quality assurance metrics.