# NES Plankton IFCB 2022: 97,026 Labeled Imaging FlowCytobot Images in 155 Classes from the Northeast US Shelf

## Facts

- Provider: Sosik Lab, Woods Hole Oceanographic Institution
- License: CC-BY-4.0 (https://huggingface.co/datasets/sosiklab/NES-plankton-IFCB-classifier-2022-dataset)
- Size: 13 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/sosiklab/NES-plankton-IFCB-classifier-2022-dataset
- Access: public, readable without an account
- Category: Environmental Science
- Source: https://ifcb-annotate.whoi.edu
- Page: https://scigantic.com/archive/50c08c7c-9f94-558a-98b8-b188b5c69286
- Keywords: Environmental Science, benchmark, Parquet, Hugging Face

## About

Training and validation images behind the NES Plankton IFCB Classifier 2022, released by the Sosik Lab at Woods Hole Oceanographic Institution (WHOI). The images are regions of interest (ROIs) recorded by the Imaging FlowCytobot (IFCB), an automated submersible flow cytometer and microscope, on the Northeast US Shelf (NES) between 2006 and 2021. There are 97,026 human-verified grayscale PNG images in 155 classes covering plankton taxa and non-plankton categories such as detritus, bubbles and fibers; labels were assigned with the IFCB Annotate tool. Each record carries the image, an integer label, the class name and the IFCB ROI persistent identifier (legacy IFCB _YYYY_DDD_HHMMSS or modern DYYYYMMDDThhmmss_IFCB form). Data are Parquet shards under data/ (8 train, 2 validation), loadable with Hugging Face datasets, pandas or pyarrow; the card includes code to export PNGs and CSV. Use it to train or benchmark plankton image classifiers and to study class imbalance in ocean imaging. Licensed CC BY 4.0 by the Sosik Lab, WHOI.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("sosiklab/NES-plankton-IFCB-classifier-2022-dataset")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `data/train-00000.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/50c08c7c-9f94-558a-98b8-b188b5c69286
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/50c08c7c-9f94-558a-98b8-b188b5c69286
- Over MCP: `get_schema_card` with id `50c08c7c-9f94-558a-98b8-b188b5c69286` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Environmental Science datasets](https://scigantic.com/topics/environmental-science)
