Plankton imaging datasets - Womack et al. manuscript

This dataset includes multiple files (plankton image datasets sourced from different imaging platforms). Each zipped file contains plankton image data from a single instrument. All data is currently part of a manuscript (Womack et al.) undergoing review. Funding: The UW38 dataset (released by us, with permission from Julie Keister at University of Washington and Deana Crouser from NOAA) was obtained with the UW ZooCam, funded by NSF Award OCE-1657992. The HBOI dataset was recorded by the Nayak Lab at Florida Atlantic University. All other raw datasets were sourced from publicly available databases - only the preprocessed sets relevant to the manuscript analysis are uploaded here. Kaggle38: Captured via the In Situ Ichthyoplankton Imaging System (ISIIS), this dataset uses shadowgraphy to provide high-resolution, non-invasive images of 38 plankton classes in their natural environment. Kaggle83: An expansion of the ISIIS shadowgraphy collection, this dataset features 83 distinct classes of plankton captured at a high tow speed of 2.5 m/s. WHOI22: Collected using the Imaging FlowCytobot (IFCB), this dataset consists of 22 classes of nano- and microplankton identified through combined fluorescence and high-resolution flow-through imaging. WHOI79: This large-scale IFCB dataset targets 79 classes, leveraging automated submersible flow cytometry to capture detailed morphological data of particles in the 10–150 µm range. ZooScan20: Derived from the ZooScan laboratory system, these 20 classes were imaged in a controlled, non-destructive scanning chamber to ensure consistent optical density and illumination. ZooScan93: A high-diversity ZooScan dataset containing 93 classes, optimized for consistent biomass and taxonomic analysis of preserved zooplankton samples. UW38: Captured by the University of Washington ZooCam, this dataset features 38 classes recorded via dark-field microscopy during vertical profiling in Hood Canal. HBOI19: This dataset uses HOLOCAM digital holography to provide 3D volumetric information and diffraction patterns for 19 classes of marine organisms. PMID23: A laboratory-based collection of 23 phytoplankton classes captured at 200x magnification using an Olympus BX53 optical microscope on preserved samples from Jiaozhou Bay. UVP60: Collected by the Underwater Vision Profiler 5 (UVP5), this dataset quantifies the vertical distribution of 60 macroscopic particle and zooplankton classes at depths up to 6000 m. ZooLake35: Focuses on 35 freshwater species from Lake Greifensee, captured using the Dual Scripps Plankton Camera (DSPC) with a dual-magnification dark-field microscope. Combined561: A massive, multi-instrument aggregate of all the above datasets, maintaining 561 distinct labels spanning various taxonomic ranks and detrital categories. Linnaeus452: A taxonomically standardized version of the Combined561 dataset, mapping 452 distinct classes to the World Register of Marine Species (WoRMS) to align labels with formal biological taxa and significant morphological categories. The preprocessed suffix indicates the dataset has been through our preprocessing script, and the 256samples suffix indicates a max of 256 randomly sampled images per class.