Open-Set Plankton Recognition

TL;DR

This work addresses open-set recognition for plankton image classification, a setting where unseen species and non-plankton artifacts appear alongside known classes. It evaluates three OSR approaches—OpenMax, ArcFace-based, and Class Anchored Clustering—on phyto- and zooplankton datasets, exploring fixed and class-specific rejection thresholds. The study introduces a new public plankton dataset pairing with existing SYKE data, and finds that CAC often yields the strongest open-set performance, while ArcFace can excel when thresholds are carefully tuned. The findings offer practical guidance for deploying OSR in automated plankton monitoring and contribute valuable public data for benchmarking future methods.

Abstract

This paper considers open-set recognition (OSR) of plankton images. Plankton include a diverse range of microscopic aquatic organisms that have an important role in marine ecosystems as primary producers and as a base of food webs. Given their sensitivity to environmental changes, fluctuations in plankton populations offer valuable information about oceans' health and climate change motivating their monitoring. Modern automatic plankton imaging devices enable the collection of large-scale plankton image datasets, facilitating species-level analysis. Plankton species recognition can be seen as an image classification task and is typically solved using deep learning-based image recognition models. However, data collection in real aquatic environments results in imaging devices capturing a variety of non-plankton particles and plankton species not present in the training set. This creates a challenging fine-grained OSR problem, characterized by subtle differences between taxonomically close plankton species. We address this challenge by conducting extensive experiments on three OSR approaches using both phyto- and zooplankton images analyzing also on the effect of the rejection thresholds for OSR. The results demonstrate that high OSR accuracy can be obtained promoting the use of these methods in operational plankton research. We have made the data publicly available to the research community.

Figures (6)

• Figure 1: Open-set plankton recognition using metric learning. Classes 1–4 represent the known classes, and ? unknown classes.
• Figure 2: Example plankton images: (a) SYKE-plankton_ZooScan_2024; (b) SYKE-plankton_IFCB_2022.
• Figure 3: Average Open-set F-scores with standard deviations from 5 experiments on the SYKE-plankton_ZooScan_2024 validation set. The plots show the average performance of the baseline models across various thresholds.
• Figure 4: Average Open-set F-scores with standard deviations from 5 experiments on the SYKE-plankton_ZooScan_2024 validation set. The plots show the average performance of the baseline models across various quantiles.
• Figure 5: Average open-set F-scores with standard deviations from 5 experiments on the SYKE-plankton_IFCB_2022 validation set. The plots show the average performance of the baseline models across various thresholds.