Use Case: FlowCam Phytoplankton Identification

Uncovering Phytoplankton with AI

Fast-track phytoplankton analysis: FlowCam AI for rapid species classification

Overview

Flowcam phytoplankton identification service uses a deep learning image recognition algorithm based on a Convolutional Neural Network (CNN) on Flowcam image data residing in the institute’s internal MongoDB database for the phytoplankton taxonomy. The output data has FAIRness characteristics following the Darwin Core standards and relevant vocabularies. With an operational environment in the iMagine-AI platform for processing images and storing the output data along with relevant guidance and documentation material, the service is available for users. Long-term (>4y) high-quality phytoplankton image datasets are also available for exploitation.

The service allows the processing of FlowCam images to determine the taxonomic composition of phytoplankton samples. The service includes setting up an operational environment for users to reuse pre-trained models, refining the AI tools for taxonomic identification, and improving the FAIRness of the full image library data as well as sampled training sets.

Main Features

  • Deploy with a combination of methods: Marketplace Inference Service to perform inference; OSCAR Inference Service: an easy-to-use system to perform inference and which will allow to track the necessary usage metrics; Train user’s own CNN
  • Authentication through EGI Check-in.
  • Source code published on GitHub to allow training on third-party infrastructure.
  • Inference AI model triggered via OSCAR
  • User’s input and output data will be kept private via OSCAR, while phytoplankton data predictions (input and output) will be shared. To keep information confidential, the user can run the model from their own server/pc. 

Explore the AI Models Underlying the Services