FastSlide

High-performance whole slide image reader for digital pathology

FastSlide is a modern C++20 library for reading whole slide images (WSI) with first-class Python support. Designed for AI/ML workflows, it provides thread-safe, efficient access to multiple slide formats.

📖 Documentation: https://docs.aifo.dev/fastslide/

📜 License: Apache 2.0

Features

🚀 High Performance - Thread-safe design
🐍 Python & C++ - Complete APIs for both languages
🔧 PyTorch Ready - Works seamlessly with DataLoader multi-worker loading
🔬 Multi-dimensional - Channels, focal planes (Z) and time points (T): full C·X·Y·Z·T selection for OME-TIFF and CZI
🖼️ Multiple images / scenes - Zeiss CZI scenes and Olympus VSI navigator/region images exposed via slide.images
📁 Multiple Formats:
- SVS (Aperio)
- QPTIFF (including mIF)
- OME-TIFF (including Z/T stacks)
- OME.ZARR (CXY, XYC only)
- MRXS (3DHISTECH, including mIF)
- iSyntax (Philips)
- Philips TIFF
- Hamamatsu NDPI (including Z stacks)
- Generic TIFF
- CZI (Zeiss, including Z/T stacks)
- Ventana (BIF)
- Olympus (VSI, including Z stacks)

QuPath Extension

FastSlide is also available as a QuPath extension, so you can open every FastSlide-supported format directly in QuPath — no coding required. The easiest way to install it is through QuPath's extension catalog (QuPath 0.6 or newer):

Open QuPath and go to Extensions → Manage extensions.
Click Manage extension catalogs → Add.

Enter the catalog URL and confirm:

https://github.com/NKI-AI/qupath-extension-catalog

Back in the extension manager, click the + next to QuPath FastSlide extension to install it. Restart QuPath if prompted.

QuPath will then read whole-slide images via FastSlide and notify you when a new version of the extension is published. Sources and manual install instructions live at NKI-AI/qupath-extension-fastslide.

Quick Start

Installation

FastSlide can be installed from a prebuilt wheel, or built from source with either Meson (simplest, integrates with pip/uv) or Bazel (hermetic, used for the official release wheels).

Option 1: Prebuilt wheel (recommended)

uv pip install fastslide

Option 2: Build from source with Meson

FastSlide is a regular Meson project. All native dependencies have wrap fallbacks, so a checkout builds standalone with nothing but a C++20 compiler, Meson (>= 1.3) and Ninja:

git clone https://github.com/NKI-AI/fastslide
cd fastslide

meson setup builddir
meson compile -C builddir

# Run the C++ test suite.
meson test -C builddir

This builds the C++ library, the fastslidetool CLI and the C API. See meson.options for the available options (e.g. -Dbuild_tool=false, -Dbuild_c_api=false, -Djpeg_decoder=jpgd).

Python package via meson-python — the Python bindings are wired up through meson-python, so the wheel builds straight from the source tree with standard Python tooling:

# Editable (development) install: compiles the native extension with Meson.
uv pip install -e .

# Or build a wheel.
uv build

pip install -e . / python -m build work the same way. The build is self-contained: all codecs are statically linked into the _fastslide extension, so the resulting wheel has no native runtime dependencies.

Option 3: Build from source with Bazel

FastSlide is a Bazel module. Builds are driven by bzlmod and we pin a specific Bazel version through .bazelversion; the recommended launcher is bazelisk, which picks up that file automatically.

git clone https://github.com/NKI-AI/fastslide
cd fastslide

# Build everything (C++ library, Python bindings, Go bindings, WASM target).
bazelisk build //...

# Run the C++ test suite.
bazelisk test //...

Building Python wheels with Bazel

Wheels are platform-specific because they bundle the native C++ extension. Each Python version has its own Bazel target: //python:fastslide_wheel_cp310 through //python:fastslide_wheel_cp314 (Python 3.10–3.14). Unlike the Meson path, Bazel can also cross-compile wheels for other platforms.

Current platform — build on the host OS/arch without cross-compilation:

# Example: Python 3.11 wheel for the machine you are on.
bazelisk build //python:fastslide_wheel_cp311

The .whl file appears under bazel-bin/python/.

Cross-compilation — build wheels for other platforms using the Zig-backed hermetic toolchains (--config=hermetic in .bazelrc):

# Example: Linux x86_64 wheel for Python 3.11, e.g. from macOS.
bazelisk build --config=hermetic --platforms=//platforms:linux_x86_64 \
  //python:fastslide_wheel_cp311

Supported platform keys: linux_x86_64, linux_arm64, darwin_x86_64, darwin_aarch64, windows_x86_64. When building for the host macOS architecture from macOS, the native toolchain is used instead of hermetic Zig.

Batch builds — tools/build_wheels.py drives Bazel for multiple platforms and Python versions and copies wheels into artifacts/wheels/:

# All supported platforms and Python versions.
python tools/build_wheels.py

# Subset, e.g. one platform and one Python tag.
python tools/build_wheels.py --platform linux_x86_64 --python cp311

# Continue after individual failures.
python tools/build_wheels.py --keep-going

Run python tools/build_wheels.py --help for the full option list.

Using FastSlide from another Bazel module

To consume FastSlide from another Bazel module, add it to your MODULE.bazel:

bazel_dep(name = "fastslide", version = "0.7.0")
git_override(
    module_name = "fastslide",
    remote = "https://github.com/NKI-AI/fastslide.git",
    commit = "<pin a recent commit SHA>",
)

and depend on @fastslide//:fastslide_lib (C++) or @fastslide//python:fastslide (Python).

Python Usage

Basic Example: Opening and Reading a Slide

import fastslide

# Open a slide using context manager (automatically closes when done)
with fastslide.FastSlide.from_file_path('slide.svs') as slide:
    # Get slide information
    print(f"Dimensions: {slide.dimensions}")  # (width, height) at level 0
    print(f"Levels: {slide.level_count}")     # Number of pyramid levels
    print(f"Resolution: {slide.mpp} µm/pixel")
    print(f"Format: {slide.format}")           # e.g., "SVS", "MRXS", "QPTIFF"

    # Read a region at full resolution (level 0)
    region = slide.read_region(
        location=(1000, 2000),  # (x, y) in level-native coordinates
        level=0,                 # pyramid level
        size=(512, 512)          # (width, height)
    ).numpy()
    # region is a numpy array: shape (512, 512, 3), dtype uint8

Example: Manual Resource Management

import fastslide

# Open a slide without context manager
slide = fastslide.FastSlide.from_file_path('slide.mrxs')

try:
    # Work with the slide
    region = slide.read_region(location=(0, 0), level=0, size=(1024, 1024)).numpy()

    # Get slide properties
    props = slide.properties
    print(f"Scanner: {props.get('scanner_model', 'Unknown')}")
    print(f"Magnification: {props.get('objective_magnification', 'N/A')}")

finally:
    # Always close the slide to release resources
    slide.close()

Example: Working with Multiple Pyramid Levels

import fastslide

with fastslide.FastSlide.from_file_path('slide.tiff') as slide:
    # Get information about all pyramid levels
    print(f"Level count: {slide.level_count}")
    print(f"Level dimensions: {slide.level_dimensions}")
    print(f"Level downsamples: {slide.level_downsamples}")

    # Read the same region at different resolutions
    location = (10000, 15000)
    size = (256, 256)

    # Full resolution (level 0)
    region_l0 = slide.read_region(location=location, level=0, size=size).numpy()

    # 4× downsampled (level 2)
    # Convert coordinates to level 2 space
    x_l2, y_l2 = slide.convert_level0_to_level_native(
        location[0], location[1], level=2
    )
    region_l2 = slide.read_region(location=(x_l2, y_l2), level=2, size=size).numpy()

    # Find best level for a specific downsample factor
    best_level = slide.get_best_level_for_downsample(8.0)
    print(f"Best level for 8× downsample: {best_level}")

Example: Multi-dimensional Reading (Channels, Z focal planes, T time points)

OME-TIFF and CZI files can store more than a single 2D plane: multiple fluorescence channels (C), a Z focal stack, and a T time series. FastSlide exposes these as a C·X·Y·Z·T hyper-volume. Each read_region selects one (z, t) plane and returns all channels of that plane; z and t default to 0, so 2D/brightfield code keeps working unchanged.

import fastslide

with fastslide.FastSlide.from_file_path('stack.ome.tiff') as slide:
    # Inspect the stack extent.
    print(f"Focal planes (Z): {slide.z_count}")
    print(f"Time points (T):  {slide.t_count}")
    print(f"Z spacing: {slide.z_spacing_um} µm")   # None if unknown
    print(f"T interval: {slide.t_interval_s} s")   # None if unknown

    # Or as a single dict.
    print(slide.get_stack_info())
    # {'z_count': 5, 't_count': 7, 'z_spacing_um': 0.5, 't_interval_s': 10.0}

    # Read the 3rd focal plane at the 2nd time point.
    region = slide.read_region(
        location=(0, 0),
        level=0,
        size=(512, 512),
        z=2,   # focal-plane index (0 = first plane)
        t=1,   # time-point index (0 = first time point)
    ).numpy()

    # Fluorescence planes carry independent channels (not RGB). The numpy
    # shape depends on the image's internal layout (see "Pixel layout" below):
    #   - interleaved / CONTIGUOUS -> (height, width, channels)  [HWC]
    #   - band-separate / SEPARATE -> (channels, height, width)  [CHW]
    for t in range(slide.t_count):
        for z in range(slide.z_count):
            img = slide.read_region((0, 0), level=0, size=(512, 512), z=z, t=t)
            # Normalize to a known layout instead of assuming one:
            arr = img.to_interleaved().numpy()  # (512, 512, channels), HWC

Per-image stacks are also available on individual images of a multi-image slide via slide.images[i].read_region(..., z=, t=) and slide.images[i].get_stack_info().

Pixel layout (interleaved vs. band-separate)

The byte layout of a returned Image is not fixed — it follows the source's internal organization, exposed via image.planar_config:

`planar_config`	Memory layout	`numpy()` shape
`PlanarConfig.CONTIGUOUS`	interleaved, HWC	`(height, width, channels)`
`PlanarConfig.SEPARATE`	band-separate,CHW	`(channels, height, width)`

Brightfield RGB is typically CONTIGUOUS; multi-channel fluorescence is typically SEPARATE. Don't assume an axis order — inspect it, or normalize:

img = slide.read_region((0, 0), level=0, size=(512, 512))

img.planar_config   # PlanarConfig.CONTIGUOUS or PlanarConfig.SEPARATE
img.is_interleaved  # True for HWC
img.is_separate     # True for CHW

# Force a specific layout (no-op + zero-copy if already in that layout).
hwc = img.to_interleaved().numpy()  # (H, W, C)
chw = img.to_separate().numpy()     # (C, H, W)

Example: Multiple Images / Scenes (Zeiss CZI, Olympus VSI)

Some files hold more than one navigable image. Zeiss CZI files expose each acquisition scene as its own image; Olympus VSI files expose a low-resolution "navigator" alongside one or more high-resolution "region" images. FastSlide surfaces these through slide.images, an indexable sequence of SlideImageViews. The top-level FastSlide accessors (dimensions, level_count, read_region, ...) always forward to the primary image, so single-image code is unaffected.

import fastslide

with fastslide.FastSlide.from_file_path('scan.czi') as slide:
    print(f"Number of images: {slide.num_images}")  # == len(slide.images)
    print(f"Image names: {slide.images.names()}")    # e.g. ['scene 0', 'scene 1']
    print(f"Primary index: {slide.images.primary_index}")

    # Iterate every image (scene) and read each independently.
    for img in slide.images:
        print(f"[{img.index}] {img.name}: {img.dimensions}, "
              f"{img.level_count} levels, Z={img.z_count}, T={img.t_count}")
        region = img.read_region(location=(0, 0), level=0, size=(512, 512)).numpy()

    # Address a specific image by index, with full level + Z/T selection.
    scene1 = slide.images[1]
    tile = scene1.read_region((0, 0), level=0, size=(1024, 1024), z=0, t=0).numpy()

    # The primary image is also directly accessible.
    primary = slide.images.primary

Each SlideImageView is a full navigator with its own pyramid (level_count, level_dimensions, level_downsamples), resolution (mpp), and Z/T stack (z_count, t_count, get_stack_info()) — so an individual scene can itself be a C·X·Y·Z·T volume.

Example: Accessing Associated Images

import fastslide
from PIL import Image

with fastslide.FastSlide.from_file_path('slide.svs') as slide:
    # Check what associated images are available
    associated = slide.associated_images
    print(f"Available images: {associated.keys()}")  # e.g., ['thumbnail', 'macro', 'label']

    # Read thumbnail (lazy loaded)
    if 'thumbnail' in associated:
        thumbnail = associated['thumbnail']  # numpy array

        # Convert to PIL Image and save
        img = Image.fromarray(thumbnail)
        img.save('thumbnail.png')

        # Get dimensions without loading
        dims = associated.get_dimensions('thumbnail')
        print(f"Thumbnail size: {dims}")

C++ Usage

#include "fastslide/slide_reader.h"
#include "fastslide/runtime/reader_registry.h"

// Create reader
auto reader = fastslide::runtime::GetGlobalRegistry()
    .CreateReader("slide.svs");

// Read region
fastslide::RegionSpec spec{
    .top_left = {1000, 2000},
    .size = {512, 512},
    .level = 0
};
auto image = reader->ReadRegion(spec);

Multi-dimensional Reading (Channels, Z focal planes, T time points)

For OME-TIFF and CZI stacks, RegionSpec::plane selects the focal plane (Z) and time point (T); both default to the first plane, so 2D reads are unaffected. Query the stack extent with GetStackInfo().

// Stack extent and physical spacing.
const fastslide::StackInfo stack = reader->GetStackInfo();
// stack.z_count, stack.t_count          : number of selectable planes (>= 1)
// stack.z_spacing_um, stack.t_interval_s : std::optional<double> (physical step)

// Read the 3rd focal plane at the 2nd time point.
fastslide::RegionSpec spec{
    .top_left = {0, 0},
    .size = {512, 512},
    .level = 0,
    .plane = {.z = 2, .t = 1},
};
auto result = reader->ReadRegion(spec);  // aifocore::Result<Image>
const fastslide::Image& image = result.value();  // check result.ok() first

// Channel memory layout follows the source and is reported per Image; do not
// assume interleaved vs. band-separate. Normalize when you need a fixed order.
const fastslide::PlanarConfig layout = image.GetPlanarConfig();
const bool interleaved = image.IsInterleaved();  // kContiguous (HWC)
const bool separate    = image.IsSeparate();     // kSeparate   (CHW)
auto hwc = image.ToInterleaved();  // zero-copy/no-op if already interleaved
auto chw = image.ToPlanar();       // zero-copy/no-op if already separate

Multiple Images / Scenes

GetImageCount(), GetImageNames() and GetImage(index) expose every navigable image (Zeiss CZI scenes, Olympus VSI navigator/region images). The reader's own ReadRegion/GetStackInfo forward to GetPrimaryImageIndex().

const int count = reader->GetImageCount();
const std::vector<std::string> names = reader->GetImageNames();

for (int i = 0; i < count; ++i) {
  auto image_or = reader->GetImage(i);  // aifocore::Result<const SlideImage*>
  if (!image_or.ok()) {
    continue;
  }
  const fastslide::SlideImage& image = *image_or.value();

  // Each image has its own name, pyramid, channels and Z/T stack.
  const std::string name = image.GetName();
  const fastslide::StackInfo stack = image.GetStackInfo();

  fastslide::RegionSpec spec{
      .top_left = {0, 0},
      .size = {512, 512},
      .level = 0,
      .plane = {.z = 0, .t = 0},
  };
  auto region = image.ReadRegion(spec);  // aifocore::Result<Image>
}

Key Features

Thread-Safe Multi-Processing

from torch.utils.data import DataLoader

# Each worker gets its own slide reader
dataloader = DataLoader(
    dataset,
    batch_size=32,
    num_workers=8,  # Safe for multi-worker loading
    shuffle=True
)

Level-Native Coordinates

FastSlide uses level-native coordinates for region reading. This is where FastSlide clearly deviates from OpenSlide, which always represents the coordinates in level 0.

# Level 0: 10000 × 8000 px (full resolution)
# Level 1: 5000 × 4000 px (2× downsample)
# Level 2: 2500 × 2000 px (4× downsample)

# Read 512×512 region from level 2 at position (100, 200)
region = slide.read_region((100, 200), level=2, size=(512, 512)).numpy()

# Convert coordinates between levels if needed
x0, y0 = slide.convert_level_native_to_level0(100, 200, level=2)
# Returns: (400, 800) - the level-0 equivalent

Documentation

📖 Complete documentation: https://docs.aifo.dev/fastslide/

Contributing

We welcome contributions. Please open an issue to discuss what you would like to change, or jump straight into a pull request.

Third-Party Components

FastSlide incorporates the following third-party software into its source:

SHA-256 implementation from sha-2 by Alain Mosnier
- Licensed under: The Unlicense or Zero Clause BSD license
- Used for: Quick hash computation compatible with OpenSlide
unordered_dense from martinus/unordered_dense by Martin Leitner-Ankerl
- Licensed under: MIT License
- Used for: Fast hashmap/hashset for spatial lookup in the Mirax format
lodepng from vandeve/lodepng by Lode Vandevenne
- Licensed under: Zlib License
- Used for: Decoding PNG in file formats and to write png in examples.
pugixml: from pugixml.org
- Licensed under: MIT License
- Used for: Parsing of XML headers
yxml: from https://dev.yorhel.nl/yxml
- Licensed under: MIT License
- Used for: Parsing of XML headers
tifffile: from cgohlke/tifffile/ by Christoph Gohlke
- Licensed under: BSD-3-Clause
- Used for: Test data files
jpeg-compressor: from richgel999/jpeg-compressor by richgel999
- Licensed under: Public domain
- Used for: Alternative JPEG decompression, required in WASM builds.
thread-pool: from bshoshany/thread-pool by Barak Shoshany
- Licensed under: MIT License
- Used for: Creating thread pool for decoding, etc.
libisyntax: from amspath/libisyntax by Pieter Valkema
- Licensed under: BSD-2 License
- Used for: iSyntax decoding
- Modifications: Library has been stripped to the minimal requirements.
jxrlib from 4creators/jxrlib by Microsoft
- Licensed under: BSD-2-Clause License
- Used for: Decoding of JPEG XR tiles in the Zeiss CZI reader
- Modifications: Library has been modified to compile with Bazel and unused files removed.

Several other libraries are used, but these are dynamically (or statically where appropriate) linked.

Citation

@software{fastslide,
  title = {FastSlide: High-performance whole slide image reader},
  author = {George Yiasemis, Rolf Harkes and Jonas Teuwen},
  year = {2025},
  url = {https://github.com/NKI-AI/fastslide}
}

Support

Documentation: https://docs.aifo.dev/fastslide/
Issues: GitHub Issues
Discussions: GitHub Discussions

License

FastSlide is licensed under the Apache License, Version 2.0.

See LICENSE for full details.

Download files

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

Source Distribution

fastslide-0.8.0.tar.gz (2.9 MB view details)

Uploaded Jul 17, 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.

fastslide-0.8.0-cp314-none-manylinux2014_x86_64.whl (2.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.14

fastslide-0.8.0-cp314-none-manylinux2014_aarch64.whl (2.2 MB view details)

Uploaded Jul 17, 2026 CPython 3.14

fastslide-0.8.0-cp314-none-macosx_11_0_arm64.whl (1.8 MB view details)

Uploaded Jul 17, 2026 CPython 3.14macOS 11.0+ ARM64

fastslide-0.8.0-cp314-none-macosx_10_9_x86_64.whl (2.1 MB view details)

Uploaded Jul 17, 2026 CPython 3.14macOS 10.9+ x86-64

fastslide-0.8.0-cp314-cp314-win_arm64.whl (1.6 MB view details)

Uploaded Jul 17, 2026 CPython 3.14Windows ARM64

fastslide-0.8.0-cp314-cp314-win_amd64.whl (1.7 MB view details)

Uploaded Jul 17, 2026 CPython 3.14Windows x86-64

fastslide-0.8.0-cp313-none-manylinux2014_x86_64.whl (2.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.13

fastslide-0.8.0-cp313-none-manylinux2014_aarch64.whl (2.2 MB view details)

Uploaded Jul 17, 2026 CPython 3.13

fastslide-0.8.0-cp313-none-macosx_11_0_arm64.whl (1.8 MB view details)

Uploaded Jul 17, 2026 CPython 3.13macOS 11.0+ ARM64

fastslide-0.8.0-cp313-none-macosx_10_9_x86_64.whl (2.1 MB view details)

Uploaded Jul 17, 2026 CPython 3.13macOS 10.9+ x86-64

fastslide-0.8.0-cp313-cp313-win_arm64.whl (1.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.13Windows ARM64

fastslide-0.8.0-cp313-cp313-win_amd64.whl (1.6 MB view details)

Uploaded Jul 17, 2026 CPython 3.13Windows x86-64

fastslide-0.8.0-cp312-none-manylinux2014_x86_64.whl (2.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.12

fastslide-0.8.0-cp312-none-manylinux2014_aarch64.whl (2.2 MB view details)

Uploaded Jul 17, 2026 CPython 3.12

fastslide-0.8.0-cp312-none-macosx_11_0_arm64.whl (1.8 MB view details)

Uploaded Jul 17, 2026 CPython 3.12macOS 11.0+ ARM64

fastslide-0.8.0-cp312-none-macosx_10_9_x86_64.whl (2.1 MB view details)

Uploaded Jul 17, 2026 CPython 3.12macOS 10.9+ x86-64

fastslide-0.8.0-cp312-cp312-win_arm64.whl (1.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.12Windows ARM64

fastslide-0.8.0-cp312-cp312-win_amd64.whl (1.6 MB view details)

Uploaded Jul 17, 2026 CPython 3.12Windows x86-64

fastslide-0.8.0-cp311-none-manylinux2014_x86_64.whl (2.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.11

fastslide-0.8.0-cp311-none-manylinux2014_aarch64.whl (2.2 MB view details)

Uploaded Jul 17, 2026 CPython 3.11

fastslide-0.8.0-cp311-none-macosx_11_0_arm64.whl (1.8 MB view details)

Uploaded Jul 17, 2026 CPython 3.11macOS 11.0+ ARM64

fastslide-0.8.0-cp311-none-macosx_10_9_x86_64.whl (2.1 MB view details)

Uploaded Jul 17, 2026 CPython 3.11macOS 10.9+ x86-64

fastslide-0.8.0-cp311-cp311-win_arm64.whl (1.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.11Windows ARM64

fastslide-0.8.0-cp311-cp311-win_amd64.whl (1.6 MB view details)

Uploaded Jul 17, 2026 CPython 3.11Windows x86-64

fastslide-0.8.0-cp310-none-manylinux2014_x86_64.whl (2.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.10

fastslide-0.8.0-cp310-none-manylinux2014_aarch64.whl (2.2 MB view details)

Uploaded Jul 17, 2026 CPython 3.10

fastslide-0.8.0-cp310-none-macosx_11_0_arm64.whl (1.8 MB view details)

Uploaded Jul 17, 2026 CPython 3.10macOS 11.0+ ARM64

fastslide-0.8.0-cp310-none-macosx_10_9_x86_64.whl (2.1 MB view details)

Uploaded Jul 17, 2026 CPython 3.10macOS 10.9+ x86-64

fastslide-0.8.0-cp310-cp310-win_arm64.whl (1.5 MB view details)

Uploaded Jul 17, 2026 CPython 3.10Windows ARM64

fastslide-0.8.0-cp310-cp310-win_amd64.whl (1.6 MB view details)

Uploaded Jul 17, 2026 CPython 3.10Windows x86-64

File details

Details for the file fastslide-0.8.0.tar.gz.

File metadata

Download URL: fastslide-0.8.0.tar.gz
Upload date: Jul 17, 2026
Size: 2.9 MB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for fastslide-0.8.0.tar.gz
Algorithm	Hash digest
SHA256	`6a58094e88c602bbb705ed8c703d7b30811f175bb659b051585cf7ae33ef896e`
MD5	`b86facdbc531ed37884498a120e29aa7`
BLAKE2b-256	`48874f79012caa500aa35e9c9803634cc7a9fd93f488ce07781851d9abb74389`

Algorithm	Hash digest
SHA256	`0e1de40fca4ce705c69ebf17b7ae094a4719a039aad6995e6c0dab8293f1eeaf`
MD5	`7eee96e9756fadd54dc4296e470145f6`
BLAKE2b-256	`eb562fc5458ec61ec003a9b66785abea204784cd87dbeaf93d61d92292c5ac29`

Algorithm	Hash digest
SHA256	`be97c40e423d765260cc153639a60b901e7a9fb0fae78924178f6d13f183a990`
MD5	`5465da9512d42f4fb32f406caa28ea53`
BLAKE2b-256	`2d5c0c16dbc037c0fe20aed4c32659c044a6bd6c6b74a3f7bfdf6b3fe4c2add5`

Algorithm	Hash digest
SHA256	`272e6ebb2e42505d4348cf700af89419796811a1cb4fbce44247eab9d7044be9`
MD5	`45749220d918b2a3153d4f170d1e0c6c`
BLAKE2b-256	`89cac3eff05be6f7774e5b5ccf518b1cf1cbe015aa4e0fb0b1717b33034c7e34`

Algorithm	Hash digest
SHA256	`d08dce1941c27443c638635b06d0f3b63941b4e308cf9c6db754c3a253847f9f`
MD5	`fa311aa9f41ec8d3d23048ce12aa5523`
BLAKE2b-256	`50a643b6ec4ef38cfb02dc2e00216ff0dc8c218b7661d6456fb99d4785932b91`

Algorithm	Hash digest
SHA256	`c4a29037a7029bc82e974b90dbcd622b7be0d94b3ed96e60522d991b834ae779`
MD5	`aa1f3af86e07c70f21dfe2accb584ce7`
BLAKE2b-256	`57ff64e74b14fe877fc483b4b757405c899222d20e2aafc3bb3db0b98e7df721`

Algorithm	Hash digest
SHA256	`384c288b539c9aabfe42ed68323527a2eba2603cd7ccbbcd2e68c8b16ffd0991`
MD5	`8bb0bb3970e261e641c00e22a6e3ff4f`
BLAKE2b-256	`d120b50fe2edf38294bc11acd17e1d0ba57ed987c33860fdd171568dbbf991a2`

Algorithm	Hash digest
SHA256	`1a2e89c6d6ce47647475a9570c910947517ea38cc957bc1fce016085420f60d7`
MD5	`a84d9e743253b85c2c6349138ac111a0`
BLAKE2b-256	`37093fcce65e9f734c6a3df268ac8b9e9dbc5739356d50a43a24e99c4041e43b`

Algorithm	Hash digest
SHA256	`1fe75a7f7db03f7b42eeb873e4c2e9408bc5f96fad7199995edac2a478efcf14`
MD5	`466900641d74f32146a944d2bf56236a`
BLAKE2b-256	`510dbe648ae9e7a8d72b4cef0a3b8e717e33b347e69549ceceee9a877acd3386`

Algorithm	Hash digest
SHA256	`db1126edeb36e599fd431a712fa6d765e9aea88450a523bad679ec19b0b84546`
MD5	`89f34f22d6bdeff101aa1be080d27efc`
BLAKE2b-256	`34183945f88a26f67d3a6fd48c60111eae67b6b2c6fd9cf600bd7642c0e3822a`

Algorithm	Hash digest
SHA256	`713fbfcda4416e62a15a7e16c551cfec17aedf1b9fcbc351052db7e3f6243015`
MD5	`9bf604845e8ca544160648b4314d6cc7`
BLAKE2b-256	`9c4648cded0665145b4fb26983446b68fb013314650004453d7cbff4802dbd26`

Algorithm	Hash digest
SHA256	`3f5b07ed57d5aa90b0ece5d8d4982f2d44ee9fbf7326953dc4562559dd4d7dbd`
MD5	`efe1fc233ca4077da549a15a32d121e0`
BLAKE2b-256	`c575e464dc7404898e94cf79bb3bbfadf07ce9b4b2e903623098148681c9a7c8`

Algorithm	Hash digest
SHA256	`f410379c85befb7bb66d6fdf9eae249d9483afdcadb57c2fc89dd0d5f6f14ae1`
MD5	`f73fae72de7ecb11d2cb5d2682d9b6dd`
BLAKE2b-256	`eb39aab1544efd34d1565b225eaa75473ab2c7b754d7fd726da11c7ee3681046`

Algorithm	Hash digest
SHA256	`6ec352c382cb192b81d96d66f6bdef0e6efb5455474546d1508dbbd88b468d99`
MD5	`f0708a81d39b6cc405c35f6fdd9a3402`
BLAKE2b-256	`924d03c99b626898333098d97cab062903543bc7955624e0a5d6ebd090245815`

Algorithm	Hash digest
SHA256	`9bb8e0443e5eb985a839b53134516ddc341501a06d2ea221a85fc04feae433ea`
MD5	`409e2bab6602ebbce79f04bdd763ffc3`
BLAKE2b-256	`3ba1a6f06e831adb89cefa411ac57da41c9db04eb6300c5a50236497eefe12e7`

Algorithm	Hash digest
SHA256	`1050938f8eab91ec91aefa0a6cc4b9676471feeff94110c6b14989d587ca7050`
MD5	`2620ed85a90c017b5a508793df067a37`
BLAKE2b-256	`f41b9f6e6406a03e811f808350c11a84b4916ad2c20488aa97a727073b538320`

Algorithm	Hash digest
SHA256	`e72ae77e684d31fedb06b0e53890a16a5179a01817ee791d469cc3def78cbb87`
MD5	`bfb5cf076bce04945e3fc01726108fdc`
BLAKE2b-256	`36b64819567efb65e9a84bb61cc20856a205f735351ce2e45633a9a87e1b72ec`

Algorithm	Hash digest
SHA256	`3f9e4244f0615de34d05c7a57c63fcba81aade803e20b2555165648f91d211dd`
MD5	`c5be660c1854c1371c77415872bb58e6`
BLAKE2b-256	`cea714610fef7a806fcba462505f474043ea6fa35ee10d8ce4e6ebc52e1e2806`

Algorithm	Hash digest
SHA256	`0b9ac9e4afbb6e749966b90b6aa47b03abb0285569b19cf49fc4701817747475`
MD5	`a1482d57244f4fdaf1ec58610de9217e`
BLAKE2b-256	`d020e2975d4257e41daee1e569be04d32f7b4570bcd83bfa5fa8d9db84ef28d1`

Algorithm	Hash digest
SHA256	`b2c3a48f2bbb50f3248df4c06c632a575c5738037fca127b134de89e4d46a3bf`
MD5	`77c75ae942dff6bf2388e877c4a7b0ed`
BLAKE2b-256	`2ad30b5b7b808755fd8fd6793f6c9528793670601e2eb1403f8553e44f6b23e0`

Algorithm	Hash digest
SHA256	`ded3fdc90b57b96cd25fb55b0be58d336da7da3c3522d6afc40265ff7bb64816`
MD5	`0fe424915805a996700f95c880e7d464`
BLAKE2b-256	`a3c47a5a85fc154a0d41f4401297a05901c64a1d19043bb0ff31053bd28e0fe0`

fastslide 0.8.0

FastSlide

Features

QuPath Extension

Quick Start

Installation

Option 1: Prebuilt wheel (recommended)

Option 2: Build from source with Meson

Option 3: Build from source with Bazel

Building Python wheels with Bazel

Using FastSlide from another Bazel module

Python Usage

Basic Example: Opening and Reading a Slide

Example: Manual Resource Management

Example: Working with Multiple Pyramid Levels

Example: Multi-dimensional Reading (Channels, Z focal planes, T time points)

Pixel layout (interleaved vs. band-separate)

Example: Multiple Images / Scenes (Zeiss CZI, Olympus VSI)

Example: Accessing Associated Images

C++ Usage

Multi-dimensional Reading (Channels, Z focal planes, T time points)

Multiple Images / Scenes

Key Features

Thread-Safe Multi-Processing

Level-Native Coordinates

Documentation

Contributing

Third-Party Components

Citation

Support

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

Algorithm	Hash digest
SHA256	`3b6c61392536302d3b23a8b7f5679c095ca104062a85fbf9ecdc7f528609e1c4`
MD5	`b935c8b1c4d4675e4889a1a41cefcd23`
BLAKE2b-256	`1605b028109d90c7e7906367e7d2bd80caddc417626f251494698ffb84774f2c`

Algorithm	Hash digest
SHA256	`b7421763329d1c18cce3dbace2848fe909bbb8b67fc878d1271f93756db62b0e`
MD5	`baeef07c5a04ea3644cb28d56f27249c`
BLAKE2b-256	`ff9fc8df208cfdc8e55a4a880c3a9ec6cc89641c2a0e84d492ee7047b544558c`

Algorithm	Hash digest
SHA256	`14415093d1f52fcad85b160035f02d0c1ecfe65d233300682e82f9ddaf34d9dd`
MD5	`4ddd995437e1df64111d51dc3e41a1d7`
BLAKE2b-256	`8eef8e3cd9f833f5725b5a78ce6803d58cc97f25c10d670230d4ec4e8da298ce`

Algorithm	Hash digest
SHA256	`a5e975c1fd7a2899b3aba522dcb5000bdd44a271f83862b3a71a6bce4cb20af3`
MD5	`a3a867638e4a1cbad600814f25bab6ce`
BLAKE2b-256	`e4872bf83d29a11124894b602f8a60aa58c141f551de97998ee22c19e4fe1568`

Algorithm	Hash digest
SHA256	`b8a27b7bf1b8f7ac138dc0d9d8e5f09f60b40cf7b6619ec44a38e14e5478001b`
MD5	`343ef2084c3c73925e62c4d49bd0257b`
BLAKE2b-256	`81b8b85efcb5a8dcdb810af75e51abdb0fda2ea7f23ae5b1a827fcca59dcf4fa`

Algorithm	Hash digest
SHA256	`f899bb1c83080688e36abac10e4656cd46c21a97503b2a4aa1a75c7d2e8df5c3`
MD5	`4efbfa6202c749348fe65775f5b9c577`
BLAKE2b-256	`2579f0097ea08f9d752e89e56fb2c32df4bc25bf05b969e54861bc7963b82fbf`

Algorithm	Hash digest
SHA256	`2292ba497d2a8caab15e082bc557e2306e93581a64098890ec0230baf98aeaaf`
MD5	`f4fbd759bb75f6cf170c5931b5e3c4ce`
BLAKE2b-256	`a7d8111ecb8398a9f9029490fea00dc1787f3770b0057774837ec4fdd8712a4e`

Algorithm	Hash digest
SHA256	`d43f30baff19d3c5219302f18159fcc868f320ffbd9cc30c3df63e57cc076220`
MD5	`dbe28704e3717d7f5d6f4f04f8626dff`
BLAKE2b-256	`4aa166ebbb0e0f58daf4919b90e16e23b3f94e60db5c7cfcfd65d99cc6e88836`

Algorithm	Hash digest
SHA256	`c4f3d63860ed3c5b807b13ecdfc9dae11cff990a4e709ad1eab5c6512c20e111`
MD5	`67195b95a704a84ac80f7612cdb3bef0`
BLAKE2b-256	`401b2b73ba4c47fdb56bdd60788cf40eeeb5f8c02fbe9cbe967ff294b6fc8c2c`

Algorithm	Hash digest
SHA256	`70074fc59d9bfba8a9f51e07a34375922c2a04f1142281dd54f792fe89632b0f`
MD5	`5d1324b7a8b08371fe3860acff6340da`
BLAKE2b-256	`ddf864119596ec4e6c6241e4eebaa3105f7b2ffd1be4fb775db104018df02891`