Fish image annotation data, is currently collected by various research, management and academic institutions globally (+100,000’s hours of deployments) with varying degrees of standardisation and limited formal collaboration or data synthesis. We present a case study of how national on-line services, developed within the NeCTAR Marine Sciences Cloud, have been used to annotate habitat images and synthesise fish annotation data sets collected using baited remote underwater stereo-video (stereo-BRUV).
In our case study we have used GlobalArchive to synthesise existing stereo-BRUV data sets. GlobalArchive is designed to be a centralised repository of aquatic ecological survey data with design principles including ease of use, secure user access, flexible data import, and the collection of any sampling and image analysis information. To easily share and synthesise data we have implemented data sharing protocols, including Open Data and synthesis Collaborations, and a spatial map to explore global datasets and filter to create a synthesis. To extract data, we have implemented a flexible querying protocol that allows relational database-like querying of any data stored as a flat file (e.g .txt file) with associated spatial metadata and sampling information.
In our case study we have used SQUIDLE+ to sample the habitat composition and complexity of images of the benthos collected using stereo-BRUV. SQUIDLE+ is an online platform designed for exploration, management and annotation of georeferenced images & video data. It provides a flexible annotation framework allowing users to work with their preferred annotation schemes. For this study, we have imported a CATAMI-based scheme that has been modified to allow the annotation of forward-facing images of the benthos (see http://catami.org and https://github.com/TimLanglois/HabitatAnnotation).
We demonstrate how the flexible querying protocol of GlobalArchive is used to combine habitat and fish annotation data to allow the synthesis of multiple data sets and a time-efficient and reproducible data analysis workflow.