{"success":true,"data":{"author":"Bastiaan Veeling et al. (from the CAMELYON16 challenge); Parquet conversion by 1aurent","bucketAccessType":"public","bucketName":"1aurent/PatchCamelyon","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-14T14:15:11.994Z","description":"PatchCamelyon (PCam) is a histopathology image classification benchmark introduced by Veeling et al. (MICCAI 2018) as a medical counterpart to CIFAR-10. It consists of 327,680 colour patches of 96 x 96 pixels extracted from H&E-stained whole-slide images of sentinel lymph node sections in the CAMELYON16 challenge, undersampled to 10x. Each patch carries a binary label marking whether its centre 32 x 32 region contains tumour tissue. Splits are 262,144 training and 32,768 each for validation and test, with no slide overlap between splits and a 50/50 class balance. This repository is a Parquet conversion of the original HDF5 release with image and label columns. As measured 2026-11-14 it holds 17 Parquet shards under data/ totalling about 7.6 GB; load with datasets (load_dataset('1aurent/PatchCamelyon')) or pyarrow and decode images with PIL. Used for benchmarking CNNs and vision transformers on pathology, rotation-equivariant networks, self-supervised pretraining and pathology foundation model evaluation. The original PCam repository releases the data under CC0, matching CAMELYON16; cite Veeling et al. 2018 and Ehteshami Bejnordi et al. 2017.","displayTitle":"PatchCamelyon (PCam): lymph node metastasis patch classification benchmark","id":"41af0e75-76af-55cb-91fc-8f9e269ebbe7","license":"CC0-1.0","licenseEvidenceUrl":"https://github.com/basveeling/pcam","objectCountEstimate":19,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1007/978-3-030-00934-2_24","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"1aurent/PatchCamelyon\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"1aurent/PatchCamelyon\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=1aurent, license:cc0-1.0, size_categories:100K<n<1M, downloads=5881"],"format":"mixed","generatedAt":"2026-11-14T14:15:20.619Z","layout":{"example":"data/test-00000-of-00002-bb04e6313f58efa0.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/test-00000-of-00002-bb04e6313f58efa0.parquet","truncatedAt":512}],"sampleObjects":["data/test-00000-of-00002-bb04e6313f58efa0.parquet","data/test-00001-of-00002-3bfa172e8818685a.parquet","data/train-00000-of-00013-4717c3cf92578c96.parquet","data/train-00001-of-00013-549914845b4273b1.parquet","data/train-00002-of-00013-a859720d3cfcebdf.parquet"],"sidecars":[{"bytes":"---\nconfigs:\n- config_name: default\n  data_files:\n  - split: train\n    path: data/train-*\n  - split: valid\n    path: data/valid-*\n  - split: test\n    path: data/test-*\ndataset_info:\n  features:\n  - name: image\n    dtype: image\n  - name: label\n    dtype: bool\n  splits:\n  - name: train\n    num_bytes: 6114271847\n    num_examples: 262144\n  - name: valid\n    num_bytes: 768518807\n    num_examples: 32768\n  - name: test\n    num_bytes: 750794905\n    num_examples: 32768\n  download_size: 7639680898\n  dataset_size: 7633585559\ntask_categories:\n- image-classification\ntags:\n- camelyon\n- biology\n- Histopathology\n- Histology\n- Digital Pathology\nsize_categories:\n- 100K<n<1M\nlicense: cc0-1.0\npaperswithcode_id: pcam\n---\n\n# PatchCamelyon (PCam)\n\n![](https://github.com/basveeling/pcam/raw/master/pcam.jpg)\n\n## Dataset Description\n\n- **Homepage**: [github.com:basveeling/pcam](https://github.com/basveeling/pcam)\n- **DOI**: https://doi.org/10.1007/978-3-030-00934-2_24\n- **Publication Date** 2018-09-26\n\n## Description\n\nThe PatchCamelyon benchmark is a new and challenging image classification dataset. It consists of 327.680 color images (96 x 96px) extracted from histopathologic scans of lymph node sections. Each image is annoted with a binary label indicating presence of metastatic tissue. PCam provides a new benchmark for machine learning models: bigger than CIFAR10, smaller than imagenet, trainable on a single GPU\n\n## Why PCam\n\nFundamental machine learning advancements are predominantly evaluated on straight-forward natural-image classification datasets. Think MNIST, CIFAR, SVHN. Medical imaging is becoming one of the major applications of ML and we believe it deserves a spot on the list of _go-to_ ML datasets. Both to challenge future work, and to steer developments into directions that are beneficial for this domain.\nWe think PCam can play a role in this. It packs the clinically-relevant task of metastasis detection into a straight-forward binary image classification task, akin to CIFAR-10 and MNIST. Models can easily be trained on a single GPU in a couple hours, and achieve competitive scores in the Camelyon16 tasks of tumor detection and WSI diagnosis. Furthermore, the balance between task-difficulty and tractability makes it a prime suspect for fundamental machine learning research on topics as active learning, model uncertainty and explainability.\n\n## Details\n\n### Numbers\n\nThe dataset is divided into a training set of 262.144 (2^18) examples, and a validation and test set both of 32.768 (2^15) examples. There is no overlap in WSIs between the splits, and all splits have a 50/50 balance between positive and negative examples.\n\n### Labeling\n\nA positive label indicates that the center 32x32px region of a patch contains at least one pixel of tumor tissue. Tumor tissue in the outer region of the patch does not influence the label. This outer region is provided to enable the design of fully-convolutional models that do not use any zero-padding, to ensure consistent behavior when applied to a whole-slide image. This is however not a requirement for the PCam benchmark.\n\n### Patch selection \n\nPCam is derived from the Camelyon16 Challenge, which contains 400 H\\&E stained WSIs of sentinel lymph node sections. The slides were acquired and digitized at 2 different centers  using a 40x objective (resultant pixel resolution of 0.243 microns). We undersample this at 10x to increase the field of view.\nWe follow the train/test split from the Camelyon16 challenge, and further hold-out 20% of the train WSIs for the validation set. To prevent selecting background patches, slides are converted to HSV, blurred, and patches filtered out if maximum pixel saturation lies below 0.07 (which was validated to not throw out tumor data in the training set).\nThe patch-based dataset is sampled by iteratively choosing a WSI and selecting a positive or negative patch with probability _p_. Patches are rejected following a stochastic hard-negative mining scheme with a small CNN, and _p_ is adjusted to retain a balance close to 50/50.\n\n## Citation\n\n```bibtex\n@ARTICLE{Veeling201","encoding":"utf8","key":"README.md","name":"README.md","size":4424,"truncated":true}],"sourceObjectKey":"data/test-00000-of-00002-bb04e6313f58efa0.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"1aurent/PatchCamelyon\")\nprint(ds)","starterSamplePath":"data/test-00000-of-00002-bb04e6313f58efa0.parquet","totalObjects":19},"summary":"PatchCamelyon (PCam) is a histopathology image classification benchmark introduced by Veeling et al. (MICCAI 2018) as a medical counterpart to CIFAR-10.","title":"PatchCamelyon (PCam): lymph node metastasis patch classification benchmark","updatedAt":"2026-11-14T17:46:29.826Z","visibility":"public","websiteUrl":"https://github.com/basveeling/pcam"}}