Use Case: Ecosystem Monitoring at EMSO Sites by Video Imagery

AI-Powered Ecosystem Monitoring for Healthy Oceans

The future of ocean monitoring: AI-powered photo and video analysis.

Overview

The European Multidisciplinary Seafloor and Water Column Observatory (EMSO) aims to explore the oceans, better understand the phenomena happening within and below them, and explain the critical role that these phenomena play in the broader Earth systems. Several of the EMSO sites capture underwater videos. The Marine Ecosystem Monitoring system has developed standards for managing and storing video imagery, and annotated images have been developed.
EMSO workflow has been set up in the iMagine AI Platform using artificial intelligence to preselect interesting images and analyze selected images for the identification of biota. Documentation and guidance about standard data management practices and the use of AI analysis pipelines for biota classification have also been developed.

Three services will be available on the iMagine AI Platform: a) fish abundance estimation, b) real-time fish detections, and c) benthic species detection.

Main Features

  • The AI model will run for inference for a pre-defined EMSO station which acts as its single user. The secondary users will be researchers who use the fish statistics produced from the inference, and citizens who watch the annotated live feed (from YouTube).
  • Video Quality Assessment Algorithm (Near Real time and Archive footage) based on DOVER will be used to flag poor/good quality video sequences for monitoring and archive evaluation.
  • The generated scientific datasets (both fish or benthic species detections and pictures) will be made publicly available at Zenodo under a CC-BY licence. 
  • The DEEPaas API should be used to infer AI models. The data preparation pipeline doesn’t need AI to run.
  • The user’s data will remain private and won’t be available to other users.