Standard Presentation (15 mins) Australian Marine Sciences Association 2022

SQUIDLE+: Laying the Foundation for Understanding Marine Imagery (#122)

Ariell Friedman 1 , Jacquomo Monk 2 , Stefan Williams 3 , Oscar Pizarro 3
  1. Greybits Engineering, North Bondi, NSW, Australia
  2. University of Tasmania, Hobart
  3. Australian Centre for Field Robotics, University of Sydney, Sydney

Squidle+ (SQ+; is a marine image data management, discovery and annotation platform underpinning the IMOS Understanding Marine Imagery (UMI) sub-facility. UMI is a national repository for annotations of marine imagery that provides tools for annotation, exploration, validation, sharing and exporting of this data and currently contains ~7.4M images, ~12K deployments, >1.1K users and >2.3M annotations (@15MAR22).

The architecture integrates image sources hosted on existing cloud-based repositories. Tools include map-based exploration interfaces, summary/reporting, advanced annotation workflows and analytics through a comprehensive API. Sharing, collaboration and release of datasets are managed through user groups with granular permissions. SQ+ expedites data delivery, eliminates the need to post-process data exports, and provides a centralised repository for annotations maintaining links to the originating imagery.

The platform offers different annotation modes (whole-frame, points, polygons) and is designed to be media-type agnostic (images, videos, large-scale mosaics). Integrated QA/QC tools enable cross-validation between human annotators and between humans and algorithms. SQ+ supports multiple vocabularies that are standardised by semantic translation tools, allowing data export in a vocabulary of a user’s choosing. This facilitates data reuse, syntheses between projects and large-scale training of machine learning algorithms.