Planktonzilla-17M Dataset

Overview

planktonzilla-17M is a large-scale, comprehensive dataset combining 17 million plankton images from all publicly available -to the best of our knowledge- labeled plankton datasets. This unified collection enables researchers to train robust deep learning models for plankton identification and classification across diverse imaging systems and oceanographic environments.

Each image includes a standardized taxonomic hierarchy (Kingdom, Phylum, Class, Order, Family, Genus, Species) mapped to the World Register of Marine Species (WoRMS), along with rich metadata including boolean indicators for plankton classification (live plankton vs. non-plankton) and living status (living organisms vs. detritus/inert), as well as geo-environmental metadata capturing collection location, depth, sampling conditions, and environmental context. Complete traceability to source datasets and original labels is preserved for full transparency.

To date, planktonzilla-17M represents the largest and most taxonomically complete plankton dataset ever assembled**, encompassing 599 distinct plankton classes across all major taxonomic groups and covering 198 known species with complete taxonomic lineage. The dataset captures organisms ranging from microplankton (20 μm) to large zooplankton (>300 μm), collected from 2008-2025 across global oceans, coastal regions, and freshwater environments.

Notes

• Part of Inria Challenge OcéanIA 🌊.
• Models trained on planktonzilla-17M: https://huggingface.co/project-oceania.
• Code for dataset and models: https://github.com/Inria-Chile/planktonzilla.
• Contribute: https://github.com/Inria-Chile/planktonzilla/issues

Citation

Please cite as:

A. G. Contreras Montanares, L. Valenzuela, L. Martí, and N. Sanchez‑Pi, Planktonzilla: Multimodal dataset and models for understanding plankton ecosystems, Inria Chile Research Center, Tech. Rep., May 2026, doi: 10.48550/arXiv.2606.00080, arXiv: 2606.00080 [cs.CV]. url: https://arxiv.org/abs/2606.00080

@techreport{contrerasmontanares:hal-05621003,
  title         = {Planktonzilla: {M}ultimodal dataset and models for understanding plankton ecosystems},
  author        = {Contreras Montanares, Alan Gerson and Valenzuela, Luis and Mart{\'i}, Luis and Sanchez-Pi, Nayat},
  year          = 2026,
  month         = {May},
  keywords      = {Explainable AI; XAI ; Plankton Classification ; CLIPS ; Multimodal Classification},
  eprinttype    = {arxiv},
  hal_id        = {hal-05621003},
  hal_version   = {v1},
  eprint        = {2606.00080},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CV},
  url           = {https://arxiv.org/abs/2606.00080},
  doi           = {10.48550/arXiv.2606.00080},
  institution   = {Inria Chile Research Center},
}

Dataset Structure

planktonzilla-17M aggregates data from the following imaging systems and oceanographic programs:

Dataset	Imaging System	Images	Classes	License	Source
Global UVP5Net [1]	UVP5	7.41M	254	CC-BY-NC-4.0	SEANOE
WHOI-Plankton [2]	Microscopy	3.56M	103	MIT	GitHub
JEDI System/OCEANS_CPICS [3]	Observatory System	1.92M	95	CC-BY-SA-4.0	JAMSTEC
ZooScanNet [4]	ZooScan	1.45M	120	CC-BY-NC-4.0	SEANOE
ZooCAMNet [5]	ZooCAM	1.29M	93	CC-BY-NC-4.0	SEANOE
UVP6Net [6]	UVP6	634K	54	CC-BY-NC-4.0	SEANOE
ISIISNet [7]	ISIIS	408K	32	CC-BY-NC-4.0	SEANOE
PlanktoScope [8]	PlanktoScope	180K	263	CC-BY-NC-4.0	SEANOE
FlowCAMNet [9]	FlowCAM	141K	38	CC-BY-NC-4.0	SEANOE
MedPlanktonSet [10]	IFCB	77.3K	139	CC-BY-NC-4.0	Zenodo
SYKE-IFCB 2022 [11]	IFCB	63.1K	50	CC-BY-NC-4.0	B2SHARE
PlanktonSet 1.0 [12]	ISIIS-2	60.7K	121	CC-BY-NC-4.0	NOAA NCEI
SYKE-ZooScan 2024 [13]	ZooScan	22.8K	20	CC-BY-NC-4.0	FairData
ZooLake [14]	Dual Scripps Camera	17.9K	35	CC-BY-4.0	EAWAG OpenData
Lensless [15]	Lensless Microscope	6.4K	10	CC-BY-4.0	IBM Box

Figure 1. Sample contribution by dataset.

Dataset Features

Each sample (row) in the dataset contains the following features:

Feature	Type	Description
image	image	PIL Image object in RGB format with variable resolution
dataset	string	Source dataset identifier (e.g., "flowcamnet", "isiisnet", "whoi-plankton")
original_label	string	Class label from the original dataset
original_path	string	Original file path in source dataset
Kingdom	string	Taxonomic Kingdom classification
Phylum	string	Taxonomic Phylum classification
Class	string	Taxonomic Class classification
Order	string	Taxonomic Order classification
Family	string	Taxonomic Family classification
Genus	string	Taxonomic Genus classification
Species	string	Taxonomic Species classification
proposed_label	string	Harmonized label across datasets
plankton	bool	Boolean indicating if the object is plankton
root_class	string	High-level image category (e.g., living, detritus, inert, artifact)
qualifier	string	Additional qualifier of the object (e.g., full_body, part, part_head, part_leg, part_skin, larvae, egg, etc. )
Latitude	float32	Geographic latitude of the observation (degrees)
Longitude	float32	Geographic longitude of the observation (degrees)
Depth_min	float32	Minimum sampling depth in meters
Depth_max	float32	Maximum sampling depth in meters
Temperature	float32	Water temperature at sampling location
Humidity	float32	Humidity of the sample
ObjID	string	Unique object identifier from the source dataset (Ecotaxa or IFCB)
wikidata_ID	string	Wikidata entity ID for the taxon (e.g., "Q3386609")
ecotaxa_ID	string	EcoTaxa taxon description(s); may contain several separated by ";"
aphia_ID	string	WoRMS AphiaID for the taxon
NCBI_ID	string	NCBI Taxonomy ID for the taxon
BOLD_ID	string	BOLD (Barcode of Life) ID for the taxon
timestamp	string	Acquisition date (YYYY-MM-DD) from WHOI/EcoTaxa

Taxonomic Standardization (WoRMS)

To ensure consistency across the diverse source datasets, we implemented a rigorous manual standardization process using the World Register of Marine Species (WoRMS).

Instead of redefining labels, we relied on the original class names provided by annotators in each source dataset and manually matched each of these labels to their corresponding taxonomic entry in WoRMS.

Before matching, the original labels were lightly curated to correct typographical errors and normalize naming variations (e.g., synonyms or inconsistent formatting). Then, for each processed label, the closest valid WoRMS record was identified, and its taxonomic lineage was used to populate the dataset fields (Kingdom to Species).

Geographic Coverage

We have added the sampling location when it is made available in the source dataset.

Figure 2. Location of samples provided by each dataset.

The following datasets did not provide an explicit geolocation of the samples. We have recovered that information manually, but have abstained from incorporating it to the dataset.

Dataset	Inferred site	Lat, Lon	Confidence
zoocamnet	Bay of Biscay shelf — PELGAS, R/V Thalassa	45.5, −2.5	high
isiisnet	Ligurian Sea off Nice — VISUFRONT	43.5, 7.55	high
medplanktonset	Gulf of Naples — LTER-MareChiara (SZN)	40.82, 14.25	high
syke_ifcb_2022	Baltic Sea — Utö station + ferrybox transects	59.78, 21.37	high
planktonset1.0	Straits of Florida — R/V F.G. Walton Smith, ISIIS-2	25.15, −80.55	high
zoolake	Lake Greifensee, Switzerland — Eawag DSPC	47.35, 8.68	high
sykezooscan2024	Baltic Sea — sub-basin unspecified (inferred)	~59.5, 21.4	low
planktoscope	diffuse / global multi-campaign reference set	—	n/a
lensless	lab/cultured — purchased culture, no field site	—	n/a

Temporal Coverage

Data collection spans from 2008 to 2025 across various oceanographic programs.

How to Use

Loading with Hugging Face Datasets

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("project-oceania/planktonzilla-17M")

# Access a sample
sample = dataset["train"][0]
image = sample["image"]
species = sample["Species"]
source = sample["dataset"]

---

Acknowledgments

We gratefully acknowledge all researchers and institutions who contributed the individual datasets that comprise Planktonzilla-17M.

This dataset represents the collaborative effort of the global oceanographic research community. We thank all scientists, taxonomists, and data curators who made their datasets publicly available to advance marine science and plankton research.

We also encourage continued community contributions to help complete and refine the remaining taxonomic annotations

References

1. A. C. Nocera, L. Stemmann, M. Babin, T. Biard, J. Coustenoble, F. Carlotti, L. Coppola, L. Courchet, L. Drago, A. Elineau, L. Guidi, H. Hauss, L. Jalabert, L. Karp‑Boss, R. Kiko, M. Laget, F. Lombard, A. McDonnell, C. Merland, S. Motreuil, T. Panaı̈otis, M. Picheral, A. Rogge, A. Waite, and J.‑O. Irisson, “A global consistent database of plankton and detritus from in situ imaging by the underwater vision profiler 5,” Earth System Science Data Discussions, vol. 2025, pp. 1–37, 2025. DOi: 10.5194/essd-2025-522. url: https://essd.copernicus.org/preprints/essd-2025-522/
2. Orenstein, E. C., Beijbom, O., Peacock, E. E., & Sosik, H. M. (2015). WHOI-Plankton - A Large Scale Fine Grained Visual Recognition Benchmark Dataset for Plankton Classification. CoRR, abs/1510.00745.
3. Yamasaki, H. (2022). Plankton image dataset from a cabled observatory system (JEDI System/OCEANS) deployed at coastal area of Oshima Island, Tokyo, Japan. JAMSTEC. https://doi.org/10.48518/00014
4. Elineau, A., Desnos, C., Jalabert, L., Olivier, M., Romagnan, J.-B., Costa Brandao, M., Lombard, F., Llopis, N., Courboulès, J., Caray-Counil, L., Serranito, B., Irisson, J.-O., Picheral, M., Gorsky, G., & Stemmann, L. (2017). ZooScanNet: plankton images captured with the ZooScan. SEANOE. https://doi.org/10.17882/55741
5. Romagnan, J.-B., Panaïotis, T., Bourriau, P., Danielou, M.-M., Doray, M., Dupuy, C., Forest, B., Grandremy, N., Huret, M., Le Mestre, S., Nowaczyk, A., Petitgas, P., Pineau, P., Rouxel, J., Tardivel, M., & Irisson, J.-O. (2024). ZooCAMNet: plankton images captured with the ZooCAM. SEANOE. https://doi.org/10.17882/101928
6. Picheral, M., Jalabert, L., Motreuil, S., Courchet, L., Carray-Counil, L., Ricour, F., Panaiotis, T., Petit, F., & Elineau, A. (2024). UVP6Net: plankton images captured with the UVP6. SEANOE. https://doi.org/10.17882/101948
7. Panaïotis, T., Caray-Counil, L., Jalabert, L., & Irisson, J.-O. (2024). ISIISNet: plankton images captured with the ISIIS (In-situ Ichthyoplankton Imaging System). SEANOE. https://doi.org/10.17882/101950
8. Zaccomer Hugo, Biteye Amadou, Caudan Corentin, Finet Thomas, Leblanc Karine, Le Bris Caroline, Le Gall Florence, Legeay Erwan, Marques Virginie, Moirand Laurine, Oddone Anna, Persat Romain, Portefin Emilie, Rigaut-Jalabert Fabienne, Thueux Manon, de Vargas Colomban, Vigneron Mathilde, Vogt Meike, Weymiens Lucien, Lombard Fabien (2025). PlanktoScope_reference : plankton images captured with the PlanktoScope. SEANOE. https://doi.org/10.17882/110078
9. Jalabert, L., Signoret, G., Caray-Counil, L., Vilain, M., Martins, E., Lombard, F., Picheral, M., & Irisson, J.-O. (2024). FlowCAMNet: plankton images captured with the FlowCAM. SEANOE. https://doi.org/10.17882/101961
10. Mente, M.S., Houliez, E., Scalco, E. et al. (2025) Annotated IFCB plankton images from the Mediterranean Sea. Sci Data 12, 1689. doi: 10.1038/s41597-025-05973-y
11. Kraft, K., Velhonoja, O., Seppälä, J., Hällfors, H., Suikkanen, S., Ylöstalo, P., Anglès, S., Kielosto, S., Kuosa, H., Lehtinen, S., Oja, J., & Tamminen, T. (2022). SYKE-plankton_IFCB_2022 [Data set]. B2SHARE v2. https://doi.org/10.23728/b2share.abf913e5a6ad47e6baa273ae0ed6617a
12. Cowen, Robert K.; Sponaugle, Su; Robinson, Kelly L.; Luo, Jessica; Guigand, Cedric (2015). PlanktonSet 1.0: Plankton imagery data collected from F.G. Walton Smith in Straits of Florida from 2014-06-03 to 2014-06-06 and used in the 2015 National Data Science Bowl (NCEI Accession 0127422). NOAA National Centers for Environmental Information. Dataset. doi: 10.7289/v5d21vjd.
13. Kareinen, J., & Skyttä, A. (2024). SYKE-plankton_ZooScan_2024 (Version 1). LUT University. https://doi.org/10.23729/fa115087-26984aa5-aedd-11e260b9694d
14. Kyathanahally, S. P., Hardeman, T., Merz, E., Bulas, T., Reyes, M., Isles, P., Pomati, F., & Baity-Jesi, M. (2021). Deep learning classification of lake zooplankton. Frontiers in Microbiology, 12. https://doi.org/10.3389/fmicb.2021.746297
15. Pastore, V. P., Zimmerman, T. G., Biswas, S. K., & Bianco, S. (2020). Annotation-free learning of plankton for classification and anomaly detection. Scientific Reports, 10(1), 12142. https://doi.org/10.1038/s41598-020-68662-3