The MERIDIAN team is proud to report our progress on MAIPL (the Marine Artificial Intelligence PLatform). MAIPL is an open-source web platform that will make AI techniques accessible to ocean researchers.It will be initially released in January 2024 as a multifaceted toolbox to apply AI on underwater acoustic data including graphical user interfaces to annotate acoustic data, and create databases. Furthermore, MAIPL will support the development of deep learning-based acoustic detectors and classifiers without needing to code. Our goal is to make AI techniques available to a large spectrum of ocean and environmental scientists.
The platform is designed to be highly modular and could be expanded to other data sources in the future, such as satellite image analysis and trajectory analysis. MAIPL leverages several support tools (such as Ketos and Specviz-React). Thanks to the cloud resources generously granted by the Digital Research Alliance of Canada, the MERIDIAN web development team are able to create the complex virtual infrastructure required to make all of the applications work together. For all the acoustics-related machine learning operations, MAIPL relies on ketos, and integrating the tools is leading to several advances by the data analytics team, such as new command-line options that allows ketos users to more easily perform the several machine learning tasks related to acoustic detection and classification. Each command (create databases, train models, run trained models, and compute evaluation metrics) will have a correspondent web-based application in MAIPL. With the support backend infrastructure in place and the annotation and model running tools already in advanced development stages, we are excited to continue developing and testing the platform in the coming months. Supported by CFI and Dalhousie University, MAIPL will be serving ocean researchers for years to come. We invite you to contact us if you are interested in staying up to date with MAIPL developments.