Taxa Identification with Machine Learning Enhanced by DNA Metabarcoding (TIMED) proposes combining DNA metabarcoding and optical machine learning in a novel interleaved way to pave the way toward automated aquatic biomonitoring. We will optimize the overall taxa identification pipeline consisting of imaging, metabarcoding, and optical machine learning blocks and develop novel state-of-the-art machine learning algorithms to three main tasks: 1) identification of taxa in the training set, 2) detection of taxa not in the training set, and 3) biomass estimation. For all of these tasks, we will consider novel ways to exploit the DNA metabarcoding output. During TIMED, we will collect a large-scale dataset suitable for advancing the proposed result. The project will be carried out at the Finnish Environment Institute and thus, the results will be directly available for the actual biomonitoring and support environmental decision-making.

2020

2024