Computational Pathology · owkin
camelyon16-features
Pre-extracted Phikon features from CAMELYON16 whole slide images for benchmarking cancer classification methods.
What camelyon16-features is
camelyon16-features is a computational pathology dataset published by Owkin. It contains numerical representations derived from the CAMELYON16 challenge, a widely used benchmark for detecting breast cancer metastases in lymph node tissue.
The source whole slide images were originally gathered by Radboud University Medical Center in Nijmegen and University Medical Center Utrecht in the Netherlands. Owkin generated this dataset by extracting slide features using the Phikon model.
By providing pre-computed embeddings, the dataset facilitates efficient benchmarking of multiple instance learning (MIL) algorithms and cancer classification workflows without the need to process raw histopathology slides directly.
How it was made
The dataset consists of embeddings extracted from CAMELYON16 whole slide images using the Phikon feature extraction model.
The resulting feature data are packaged in Parquet format and can be accessed using data tools such as Polars, Dask, and the Hugging Face Datasets library.
Key facts
- Publisher
- Owkin
- Feature extractor
- Phikon
- Original data contributors
- Radboud University Medical Center and University Medical Center Utrecht
- Domain
- Breast cancer lymph node metastasis detection
- Format
- Parquet
- Size category
- n < 1K
What you can use it for
- Multiple instance learning benchmarking
- Evaluating and comparing multiple instance learning (MIL) models on a standard breast cancer lymph node metastasis detection task.
- Cancer image classification
- Training downstream classification algorithms on pre-extracted feature vectors without handling high-resolution whole slide images.
Explore the data
The schema, real rows and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from owkin/camelyon16-features at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 130 rows of default/Phikon_test
features
list- min
- 968
- median
- 1,000
- mean
- 999.75
- max
- 1,000
label
int- min
- 0
- median
- 0
- mean
- 0.38
- max
- 1
Strengths and limits
Good at
- Enables rapid testing of multiple instance learning methods without requiring extensive compute for whole slide feature extraction.
- Embeddings are generated with the openly available Phikon pathology model.
- Provided in Parquet format for compatibility with tools such as Polars, Dask, and Datasets.
Watch out for
- Contains pre-extracted feature vectors rather than the original whole slide histopathology images.
- Representations are constrained to the feature space produced by the Phikon encoder.
Questions
Can I look inside camelyon16-features without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser and a SQL filter. Nothing is downloaded to your machine.
What licence is camelyon16-features under?
Its authors declare other. Read the terms on the original page before reusing the data.
Which model was used to extract the features?
The features were generated using Phikon, an openly available model on Hugging Face.
What is the clinical focus of the underlying dataset?
The data originate from the CAMELYON16 challenge, which focuses on breast cancer lymph node metastasis detection.
What file format is used for the dataset?
The features are stored and distributed in Parquet format.
Written from camelyon16-features's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.