dlup - Deep Learning Utilities for Pathology

dlup provides tools to work with large whole-slide images (WSIs) for computational pathology: tiled reading across multiple backends, tiling datasets, geometry/annotation handling, and pyramidal TIFF writing. Performance critical pieces are implemented in C++20 and exposed to Python via nanobind.

This repository is automatically synced from the AI for Oncology monorepo. It builds both with Bazel (via bzlmod) and with Meson (used to produce the Python wheels).

Features

Multiple slide backends: FastSlide, OpenSlide, tifffile, DeepZoom and remote/SlideScore
Tiling datasets: grid-based tile extraction with masks, annotations and metadata
Geometry: fast C++ polygons/points/boxes, marching squares, and Shapely interop
Annotations: import/export GeoJSON, HALO, SlideScore and dlup XML
Writers: pyramidal (libtiff-backed) and tifffile TIFF writers
Background estimation: native foreground/background masking

Installation (Python)

pip install dlup

# To enable the OpenSlide backend (bundled prebuilt libopenslide):
pip install "dlup[openslide]"

import dlup
from dlup import SlideImage

slide = SlideImage.from_file_path("slide.svs")
region = slide.read_region((0, 0), 0, (512, 512))

Building from source

Bazel

# Build the C++ libraries and Python package (native extensions)
bazelisk build //:dlup

# Run the Python test suite
bazelisk test //tests/python:tests

Meson (wheels)

# Build a wheel + sdist for the current interpreter
uv build

The Meson build fetches every native dependency through subprojects/*.wrap (aifocore, Boost.Geometry, libtiff, nanobind, ...) and links them statically, producing self-contained extensions. The pure-Python runtime dependencies (FastSlide, fim, NumPy, Shapely, ...) are installed by pip from PyPI.

Development install

For day-to-day development you can work either through the Python (uv/Meson) toolchain or through Bazel.

uv (editable, Meson backend)

dlup is a compiled (nanobind) package, so the editable install uses meson-python's rebuild-on-import hook. That hook re-invokes meson/ninja at import time, which means it must NOT be installed with build isolation — with isolation the build tools live in a throwaway environment that is deleted right after install, and the first import dlup fails with FileNotFoundError: .../bin/ninja.

Install the build tools into your environment first, then pass --no-build-isolation (with build isolation, the backend is not visible to the editable build and you get ModuleNotFoundError: No module named 'mesonpy'):

# Install the build backend + tools into the active venv FIRST.
# NB: the PyPI package is `meson-python`; it provides the `mesonpy` module that
# uv's error hint refers to (there is no separate `mesonpy` distribution).
uv pip install meson-python meson ninja nanobind

# Editable install without build isolation (so the backend + rebuild hook's
# `ninja` are found in the venv, not a deleted temp build env).
uv pip install -e . --no-build-isolation

With this setup, pure-Python edits are picked up live and C++ changes trigger a ninja rebuild on the next import dlup.

If you do not need an editable install, building and installing a regular wheel avoids the rebuild hook entirely:

uv build --wheel && uv pip install --find-links dist --force-reinstall dlup

Bazel

Bazel needs no separate install step — it builds the native extensions and resolves the Python dependencies hermetically:

# Build everything (C++ libraries + native extensions + Python package)
bazelisk build //...

# Run the Python test suite
bazelisk test //tests/python:tests

License

Apache 2.0 - see LICENSE for details.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

dlup-0.9.3.tar.gz (1.3 MB view details)

Uploaded Jul 21, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

dlup-0.9.3-cp314-none-manylinux2014_x86_64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.14

dlup-0.9.3-cp314-none-manylinux2014_aarch64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.14

dlup-0.9.3-cp314-none-macosx_11_0_arm64.whl (1.5 MB view details)

Uploaded Jul 21, 2026 CPython 3.14macOS 11.0+ ARM64

dlup-0.9.3-cp314-none-macosx_10_9_x86_64.whl (1.7 MB view details)

Uploaded Jul 21, 2026 CPython 3.14macOS 10.9+ x86-64

dlup-0.9.3-cp314-cp314-win_arm64.whl (1.2 MB view details)

Uploaded Jul 21, 2026 CPython 3.14Windows ARM64

dlup-0.9.3-cp314-cp314-win_amd64.whl (1.3 MB view details)

Uploaded Jul 21, 2026 CPython 3.14Windows x86-64

dlup-0.9.3-cp313-none-manylinux2014_x86_64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.13

dlup-0.9.3-cp313-none-manylinux2014_aarch64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.13

dlup-0.9.3-cp313-none-macosx_11_0_arm64.whl (1.5 MB view details)

Uploaded Jul 21, 2026 CPython 3.13macOS 11.0+ ARM64

dlup-0.9.3-cp313-none-macosx_10_9_x86_64.whl (1.7 MB view details)

Uploaded Jul 21, 2026 CPython 3.13macOS 10.9+ x86-64

dlup-0.9.3-cp313-cp313-win_arm64.whl (1.2 MB view details)

Uploaded Jul 21, 2026 CPython 3.13Windows ARM64

dlup-0.9.3-cp313-cp313-win_amd64.whl (1.3 MB view details)

Uploaded Jul 21, 2026 CPython 3.13Windows x86-64

dlup-0.9.3-cp312-none-manylinux2014_x86_64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.12

dlup-0.9.3-cp312-none-manylinux2014_aarch64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.12

dlup-0.9.3-cp312-none-macosx_11_0_arm64.whl (1.5 MB view details)

Uploaded Jul 21, 2026 CPython 3.12macOS 11.0+ ARM64

dlup-0.9.3-cp312-none-macosx_10_9_x86_64.whl (1.7 MB view details)

Uploaded Jul 21, 2026 CPython 3.12macOS 10.9+ x86-64

dlup-0.9.3-cp312-cp312-win_arm64.whl (1.2 MB view details)

Uploaded Jul 21, 2026 CPython 3.12Windows ARM64

dlup-0.9.3-cp312-cp312-win_amd64.whl (1.3 MB view details)

Uploaded Jul 21, 2026 CPython 3.12Windows x86-64

dlup-0.9.3-cp311-none-manylinux2014_x86_64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.11

dlup-0.9.3-cp311-none-manylinux2014_aarch64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.11

dlup-0.9.3-cp311-none-macosx_11_0_arm64.whl (1.5 MB view details)

Uploaded Jul 21, 2026 CPython 3.11macOS 11.0+ ARM64

dlup-0.9.3-cp311-none-macosx_10_9_x86_64.whl (1.7 MB view details)

Uploaded Jul 21, 2026 CPython 3.11macOS 10.9+ x86-64

dlup-0.9.3-cp311-cp311-win_arm64.whl (1.2 MB view details)

Uploaded Jul 21, 2026 CPython 3.11Windows ARM64

dlup-0.9.3-cp311-cp311-win_amd64.whl (1.3 MB view details)

Uploaded Jul 21, 2026 CPython 3.11Windows x86-64

dlup-0.9.3-cp310-none-manylinux2014_x86_64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.10

dlup-0.9.3-cp310-none-manylinux2014_aarch64.whl (2.1 MB view details)

Uploaded Jul 21, 2026 CPython 3.10

dlup-0.9.3-cp310-none-macosx_11_0_arm64.whl (1.5 MB view details)

Uploaded Jul 21, 2026 CPython 3.10macOS 11.0+ ARM64

dlup-0.9.3-cp310-none-macosx_10_9_x86_64.whl (1.7 MB view details)

Uploaded Jul 21, 2026 CPython 3.10macOS 10.9+ x86-64

dlup-0.9.3-cp310-cp310-win_arm64.whl (1.2 MB view details)

Uploaded Jul 21, 2026 CPython 3.10Windows ARM64

dlup-0.9.3-cp310-cp310-win_amd64.whl (1.3 MB view details)

Uploaded Jul 21, 2026 CPython 3.10Windows x86-64

File details

Details for the file dlup-0.9.3.tar.gz.

File metadata

Download URL: dlup-0.9.3.tar.gz
Upload date: Jul 21, 2026
Size: 1.3 MB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.14

File hashes

Hashes for dlup-0.9.3.tar.gz
Algorithm	Hash digest
SHA256	`7c2b6dca83cb242dea222d71641a5f69b27cb017254d95fc911bdd339c59bd5f`
MD5	`1b42b411fc864982a71e8d53e704222a`
BLAKE2b-256	`d498692846c378b391d820552f5badc8f10c64534579f83b00e2d708acb4102e`

Algorithm	Hash digest
SHA256	`d0d30b38f649a6bf90944ad8dcc75fcf48c94f2568d6dc057f9b604d488f2e08`
MD5	`8ae6d8d0fb7531f622f5edc0a053f195`
BLAKE2b-256	`d5be28d6db3738a74fd59717649c105942bbf2f2122cd3b080e0e1526f5d0200`

Algorithm	Hash digest
SHA256	`10e1376e3a0db2263b3f8c24b2a8d7f9ac76eabf76fd6f84fb77c24529b6ee73`
MD5	`5ec9c0c95025fc7838188d0bfbbca546`
BLAKE2b-256	`f491621ca9153ec771509034348084703ab9294d4df46a7151b8c8f3e408da47`

Algorithm	Hash digest
SHA256	`d09a0d61f7a07bc0105a900f67908a7dce0e6a898b5b2dee2bad12d9e1bfe8d6`
MD5	`a09fa497acd6652fe2120752e963ab47`
BLAKE2b-256	`f13cb73f4022f7232a0e7ed07a7542bb8d6ebda87b1e05c768cb327440174fe6`

Algorithm	Hash digest
SHA256	`d1267052ffdc69c84785a1574601c654c649fb74c323a82ef106e1a10f98326d`
MD5	`4e5cabc253c1978a62a7e732293bff86`
BLAKE2b-256	`5646428eee446db6253f2b5ab3ccfd8962fe35e33905ce8df79a53eb0365b5eb`

Algorithm	Hash digest
SHA256	`9a0a40f2396e9d1c8bfdd206d561fcd8da7246e3f7ea145d0e21994907ef52b3`
MD5	`46f3f14f015b4af8bab302ae703568a4`
BLAKE2b-256	`543b072550f4a5970c2b42473adb888ddcfd799dfb585764b8dfc373ea7e8eb1`

Algorithm	Hash digest
SHA256	`51cad0e5cb854e58d6ba95cbca9d52507aa88c818dce53e0f2e68bd10a19a31a`
MD5	`9b27bee01bb115fdc761cc422b70a232`
BLAKE2b-256	`70e5f5673031a7d099645be116d47073e056374e872924b7fef13cc4363fa221`

Algorithm	Hash digest
SHA256	`1369ddd74eabafa029dfd93c591e213bcc73a7899a319da519cfdcaab0d6cd9e`
MD5	`a90e146e7045a5b6eac8675726dd54c3`
BLAKE2b-256	`36f6fdfa08230444f604e33efc64ef8eb9ca8ebbdc4f9307c8c33e9005431e04`

Algorithm	Hash digest
SHA256	`86efa082eec78e352e7ea18738f78195e89cf14f83505e4f485c4fe125005caf`
MD5	`6cfe64f3730cfcc7f0522ecf49c749e8`
BLAKE2b-256	`d2b9249595a010efac14a8395854a5655cbc29fa7000028902e71d0af88cbb77`

Algorithm	Hash digest
SHA256	`3d6e01647c72be6503d34f2df6736b188386080df4f41ab609b4a15351b5b6d8`
MD5	`166e56c9ac810b26987e6568999e1995`
BLAKE2b-256	`23aed1d44e07977ab3e44d744b0c1bfa65e3c34f8ea98ba3ca352b6377f66a39`

Algorithm	Hash digest
SHA256	`92a3d155bdb9b8b9ad93cb58ba64ec47fc83b0aa1d57886a23bd741ad8552d38`
MD5	`f8686c2c2e13d315fe18f65a7a252b07`
BLAKE2b-256	`9c7d9193c4fad8e0722be6af79a1a1d40d785a6aecdab1b93aaf82a44b8fca2c`

Algorithm	Hash digest
SHA256	`e658e716688e8ace246b0324310e255f7740361c5997c66fbee25e3f15bd5fe4`
MD5	`d60afb938617a2458624662f4b34eefb`
BLAKE2b-256	`102fe554b30e697f5e6b46492291908ef208fa91088471c52ee3c13a905c4746`

Algorithm	Hash digest
SHA256	`f5f4a9227782c434ba653e9179e28a9fc489ac825c39eb8fa5c038a1bb3091d3`
MD5	`a0c95134d7ac9d0b67f3069053024cfe`
BLAKE2b-256	`cf670fb411b39dbd7ea627d2aceb5689d3a1f6f0cc5a1b0779b167b9227e257e`

Algorithm	Hash digest
SHA256	`fd9e59c20e4e48acf946b028bf7a67802b9254d533f4dff5e3b665d55cb6d27f`
MD5	`18d66b9997b0be7b829b0ef0de49df4e`
BLAKE2b-256	`444681d28ecd5f278e0b7ae75ca02924c94aba3405e1e8642aa253c5fbe50bf5`

Algorithm	Hash digest
SHA256	`bd36ca19cf0cc14c5f9c90825a4b0540c3101f493999a7f9176e60294d02487e`
MD5	`300dbdaba6e8cdce967a08ebb1b93e96`
BLAKE2b-256	`d89c0c9f386df11f5307a41870c83ac85ea7cf78dce06ab52745ef84871619e8`

Algorithm	Hash digest
SHA256	`0617d09b735d5576943379c1b0a615ea09623b8ccc4e9346637a2319e28e6de1`
MD5	`b1cf905e6fa531470e299903448471bc`
BLAKE2b-256	`c36641b39872a6c773b4dbea98d08457062c259af8dfb385df831176f1eaa1be`

Algorithm	Hash digest
SHA256	`a6d157a6ac19d37a6edc695a36ece2f193dc2d21c21f40ac0c275477935e9de2`
MD5	`f5738895d02f67a68069e3b2e87d5dfe`
BLAKE2b-256	`aa6618d960ce9fa6c21d043b7c1fc894594af9d19405606f0804c363639e7b6f`

Algorithm	Hash digest
SHA256	`91b171ad88060b9bb4bed5982022765ac8cf922b93c5344a0f763deb3a130c33`
MD5	`6b07fef749154872de1ecc1f878ed9f8`
BLAKE2b-256	`b22993b737c720083a0d46eb0580f6c5929b0b4185542659ad0c588b80d08240`

Algorithm	Hash digest
SHA256	`a037d5f704ac237d2464c39a3ec889a1703eda9c98217dd8fef208f752d9ef6b`
MD5	`8c8c917754279f6e7776e7ee4445a400`
BLAKE2b-256	`90408cb573b1f5daeb65a7da9a1b2e5245db58a94192cf2ce5174a6412c4c4f9`

Algorithm	Hash digest
SHA256	`1174b1364c1e6ce29d0aef0dfe82e911ca43b63d85d864a08656393a9e7c8d9e`
MD5	`36471485f398be333e75967ccce0c820`
BLAKE2b-256	`914c05ceb254a92f40e5bf971656672687b5c535c509614f4333e356768ea18e`

Algorithm	Hash digest
SHA256	`57da934d1769fb595358f09bb3ec97b3adc06474be46ff2c69bdb650279d45eb`
MD5	`b45db4411a68282427f0714f5f0c6413`
BLAKE2b-256	`154917655cf5551e4f1750cdad17a6ffef4a80f21dc886e92186e86e0672aff8`

dlup 0.9.3

dlup - Deep Learning Utilities for Pathology

Features

Installation (Python)

Building from source

Bazel

Meson (wheels)

Development install

uv (editable, Meson backend)

Bazel

License

Project links

Key dates

1 maintainer

Credits

GitHub Statistics

License

Requires

Provides Extra

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

Algorithm	Hash digest
SHA256	`0537330fec0986a97423bd8b96c33bc8d8c143223b9fd9d6ee9f3ed94c1414e4`
MD5	`f83f7bc08d263f5194126e0883eefa3f`
BLAKE2b-256	`bc6a94c3e87d2e8bcf3e777f4224ac269253c37f8f1122d1310f023b27ad4253`

Algorithm	Hash digest
SHA256	`3fbdeca55ce8f1b147ce2cc445ca418dcdd6f74dedfc059d5b9764dad42e796c`
MD5	`0f629c697c31068daeea8f1d23711d80`
BLAKE2b-256	`6ca257516dc4f65c6aaae2466cde903576b79ecb78049bf27a9f1a4bc431dc20`

Algorithm	Hash digest
SHA256	`da2a1218dd6f9b74b298ab8c45e78a8840b2ea1c5c45a67a1ead2344e85b96fe`
MD5	`9f43af4871bf722d714bf6f7bd12d8a2`
BLAKE2b-256	`ddc8305d5190240f56d7c9d77f630293bf6033e4de97f8470b0d0e6e223c436c`

Algorithm	Hash digest
SHA256	`22458611adb4140ba13a7bd65014246090f0801c7283b244e819d0d6f1841ed1`
MD5	`5ab7f994d9da5809e93d6d6bce196ba0`
BLAKE2b-256	`ac471166b1cb7b41372081e11d21a0e3aa27dba11467a80ce523ed69f48dca64`

Algorithm	Hash digest
SHA256	`e71a475209fa349cbaf1ec2ce9dc77cc75a6b9884f06caef01c1ef6797433bd7`
MD5	`cbd68a4ded8e6b7c23d81410d723d693`
BLAKE2b-256	`24da77f21eb27a817c61832feb0f73f20d05ebe48fdb31aab20cda11d17c7389`

Algorithm	Hash digest
SHA256	`f4ec23acc48c2683a1ca01c026f29062012c0f05bc66c798c7c1663ca4b27ac1`
MD5	`d6c06d8137ddf39f370e1502bf4dfdf2`
BLAKE2b-256	`cb1837e1132089cfca666a78b572d8a025a46135df5fccbb2e1c764a8c6eba05`

Algorithm	Hash digest
SHA256	`d1d7d8d93d628d144b6cf1444d3c080b3402316b92d784870eed2f77d275247f`
MD5	`c78a06caf04595d77cbb62ee4a9254c5`
BLAKE2b-256	`b8ec81d17827d7f5e1f151a9c569f9c64c68e200db7b6b3fdf03d60ca16bf99d`

Algorithm	Hash digest
SHA256	`950219b3dc7fa36ec762efdd3983b44c4f0d846ee902f7ebd1efe3c8827b7e9c`
MD5	`400eb1334e67fd274d7c5e839d5b0bf3`
BLAKE2b-256	`34de5ac5df3e60c645083f3d17409925f6a5b3a3b588a24cb86165f2c75b7107`

Algorithm	Hash digest
SHA256	`1967d07f1521834406fff306176ab757cf8f0cd4b1134f94fe5c6752f0294855`
MD5	`a53611eeb4a90da1f027917173e74369`
BLAKE2b-256	`e43acaa38178b30386f65563fc4dfbb245a05aca853e73a30f825916a32134af`

Algorithm	Hash digest
SHA256	`4b1019880b8cfbb6934263520627d78dd813772e8eee17ae1cbc3457f8fee12b`
MD5	`9bddea78a684ae4e030ee55a7e542d2d`
BLAKE2b-256	`5c2925781e3de7a640b4f6d2698df779d620632e117aedc229745b25fbf30897`