Presentation Information
[eHvadpaV-03]Squidle+ and SquidJam: An Explorable, FAIR Digital Twin of the Ocean for AI-Ready Marine Imagery
*Dhugal John Lindsay1 (1. JAMSTEC (Japan))
The rapid growth of marine imaging — from towed cameras, ROVs and AUVs to citizen-science divers — has outpaced our ability to turn pixels into usable knowledge. Squidle+ (squidle.org) is a web-based, collaborative platform providing end-to-end management, discovery, annotation, and reuse of marine imagery, regardless of collection method, ecosystem, or data format. Rather than duplicating imagery, it links directly to existing cloud storage repositories, decoupling collection from delivery and enabling automated, set-and-forget ingestion.
A core strength is Squidle+'s map-based exploration interface: users can browse a world map of surveys, zoom into individual deployment tracks, and click through to seafloor imagery and photomosaics stitched from thousands of stereo pairs, all georeferenced and contextualised with annotation summaries. In this sense, Squidle+ already functions as an early, imagery-based digital twin of the seafloor — a persistent, explorable, spatially-indexed representation of real ocean environments that anyone can query and revisit.
Beyond exploration, Squidle+'s annotation framework is media-agnostic, spanning stills, video, and mosaics through one consistent interface, with QA/QC tools enabling human-human and human-algorithm cross-validation. Its vocabulary translation framework cross-walks between disparate labelling schemes (CATAMI, WoRMS, CAAB) via direct mapping, hierarchical traversal, and semantic matching, enabling cross-project synthesis, large-scale ML training, and national reporting. ML algorithms are treated as first-class "users," connecting researchers directly to validated training data through the same sharing framework used for human collaborators. Squidle+ now holds over 10.2 million images from 22,107 deployments, underpinning 50+ publications, including in Nature and Nature Ecology & Evolution.
This talk will also introduce SquidJam, a complementary tool for streamlining metadata ingestion and validation from historical survey platforms, flagging sensor gaps and timing errors before they propagate downstream. Together, Squidle+ and SquidJam show how open, FAIR imagery infrastructure can serve as both a practical building block and an early realisation of ocean digital twins.
A core strength is Squidle+'s map-based exploration interface: users can browse a world map of surveys, zoom into individual deployment tracks, and click through to seafloor imagery and photomosaics stitched from thousands of stereo pairs, all georeferenced and contextualised with annotation summaries. In this sense, Squidle+ already functions as an early, imagery-based digital twin of the seafloor — a persistent, explorable, spatially-indexed representation of real ocean environments that anyone can query and revisit.
Beyond exploration, Squidle+'s annotation framework is media-agnostic, spanning stills, video, and mosaics through one consistent interface, with QA/QC tools enabling human-human and human-algorithm cross-validation. Its vocabulary translation framework cross-walks between disparate labelling schemes (CATAMI, WoRMS, CAAB) via direct mapping, hierarchical traversal, and semantic matching, enabling cross-project synthesis, large-scale ML training, and national reporting. ML algorithms are treated as first-class "users," connecting researchers directly to validated training data through the same sharing framework used for human collaborators. Squidle+ now holds over 10.2 million images from 22,107 deployments, underpinning 50+ publications, including in Nature and Nature Ecology & Evolution.
This talk will also introduce SquidJam, a complementary tool for streamlining metadata ingestion and validation from historical survey platforms, flagging sensor gaps and timing errors before they propagate downstream. Together, Squidle+ and SquidJam show how open, FAIR imagery infrastructure can serve as both a practical building block and an early realisation of ocean digital twins.
Comment
To browse or post comments, you must log in.Log in