RadFiled3D

Tests

This Repository contains the file format and API according to the Paper: "RadField3D: A Data Generator and Data Format for Deep Learning in Radiation-Protection Dosimetry for Medical Applications".

The aim of this library is, to provide a simple to use API for a structured, binary file format, that can store all relevant information from a three dimensional radiation field calculated by applications that use algorithms like Monte-Carlo radiation transport simulations. Such a binary file format is useful, when one needs to process a huge amount of radiation field files like when training a neural network. With that use-case in mind, RadFiled3D also provides a python interface with a pyTorch integration. In order to directly iterate a dataset generated with the RadField3D tool, just jump to the section RadField3D Datasets.

🌟 Why Use RadFiled3D

Efficient Storage: Structured, binary file format for storing large amounts of radiation field data.
Easy Integration: Simple API for C++ and Python with pyTorch support.
High Performance: Optimized for fast data access and manipulation.
Versatile: Supports both Cartesian and Polar coordinate systems.
Extensible: Easily extendable to include additional metadata and data types.

Building and Installing
Getting Started
Field Structure
Dependencies

Building and Installing

Installing from PyPi

Prebuilt versions of this module for python 3.11, 3.12 and 3.13 for Windows and most Linuxsystems can be installed directly by using pip.

pip install RadFiled3D

Installing from Source

You can build and install this library and python module from source by using CMake and a C++ compiler. The CMake Project will be built automatically, but will take some time.

Prerequisites

C++ Compiler
- g++ or clang for Linux
- MSVC or clang from Visual Studio 2022 for Windows
CMake >= 3.30
Python >= 3.11

CMake

In order to use the module directly from another C++ Project, you can integrate it by adding the local location of this repository via add_submodule() and then link against the target libRadFiled3D. All classes are then available from the namespace RadFiled3D. Check the Example or the First Test File as a first reference.

Python

The Python package is built with scikit-build-core, the standard PEP 517 backend for CMake projects. It drives the CMake/pybind11 build automatically; no setup.py is required. CMake and Ninja are provisioned by the build backend if they are not already present.

Installing locally

python -m pip install .

Building a wheel

python -m build --wheel

The package version is taken from the release tag (GITHUB_REF / CI_COMMIT_TAG) at build time and falls back to 0.0.0 for non-release builds.

Getting Started

Disclaimer: Not all methods support keyword arguments as they need to be defined manually in the bindings. For some methods like add_layer or the Metadata methods those are implemented.

From Python

Simple example on how to create and store a radiation field. Find more in the example file: Example

from RadFiled3D.RadFiled3D import vec3, CartesianRadiationField, DType
from RadFiled3D.utils import FieldStore, StoreVersion
from RadFiled3D.metadata.v1 import Metadata


# Creating a cartesian radiation field
field = CartesianRadiationField(vec3(2.5, 2.5, 2.5), vec3(0.05, 0.05, 0.05))
# defining a channel and a layer on it
field.add_channel("channel1").add_layer("layer1", "unit1", DType.FLOAT32)

# accessing the voxels by using numpy arrays
array = field.get_channel("channel1").get_layer_as_ndarray("layer1")
assert array.shape == (50, 50, 50, 1)

# modify voxels content by using numpy array as no data is copied, just referenced
array[2:5, 2:5, 2:5] = 2.0

# addressing a voxel by providing a point in space
voxel = field.get_channel("channel1").get_voxel_by_coord("layer1", 0.1, 2.4, 2.1)

# Store changes to a file
metadata = Metadata.default()
FieldStore.store(field, metadata, "test01.rf3", StoreVersion.V1)

# load data
field2 = FieldStore.load("test01.rf3")
metadata2 = FieldStore.load_metadata("test01.rf3")

Integrating with pyTorch

RadFiled3D comes with a submodule at RadFiled3D.pytorch. This module provides some dataset classes to support the usage. Datasets can be loaded from folders or .zip-Files.

from RadFiled3D.pytorch.datasets import MetadataLoadMode
from RadFiled3D.pytorch.datasets.cartesian import CartesianFieldSingleLayerDataset
from RadFiled3D.pytorch import DataLoaderBuilder
from RadFiled3D.pytorch.helpers import RadiationFieldHelper
from RadFiled3D.RadFiled3D import VoxelGrid
from torch import Tensor
from RadFiled3D.metadata.v1 import Metadata
from RadFiled3D.pytorch.types import TrainingInputData, DirectionalInput


# Extend one of the provided dataset classes to match the output to the current needs
class MyLayerDataset(CartesianFieldSingleLayerDataset):
    def __getitem____(self, idx: int) -> TrainingInputData:
        layer, metadata = super().__getitem__(idx)
        tube_dir = metadata.get_header().simulation.tube.radiation_direction
        tube_pos = metadata.get_header().simulation.tube.radiation_origin
        # transform the layers data to a tensor
        return TrainingInputData(
            input=DirectionalInput(
                direction=torch.tensor([tube_dir.x, tube_dir.y, tube_dir.z]),
                origin=torch.tensor([tube_pos.x, tube_pos.y, tube_pos.z]),
                spectrum=None
            )
            ground_truth=RadiationFieldHelper.load_tensor_from_layer(layer)
        )


def finalize_dataset(dataset: MyLayerDataset)
    dataset.set_channel_and_layer("test_channel", "test_layer")
    dataset.metadata_load_mode = MetadataLoadMode.HEADER

# Pass the dataset class and other options to the DataLoaderBuilder
builder = DataLoaderBuilder(
    "./test_dataset.zip",
    train_ratio=0.7,
    val_ratio=0.15,
    test_ratio=0.15,
    dataset_class=MyLayerDataset,
    on_dataset_created=finalize_dataset     # Optional: provide a finalizer to perform configuration of the dataset once it was created by the builder
)

# Build the training dataset
train_dl = builder.build_train_dataloader(
    batch_size=8,
    shuffle=True,
    worker_count=4
)

# iterate over the dataset
for field, metadata in train_dl:
    pass

Direct integration with RadField3D datasets

Directly iterate RadField3D datasets either by loading whole fields or iterating each voxel independently. The dataset classes will return pyTorch compatible NamedTuples, that preserve the structure of the raw radiation fields and layers.

from RadField3D.pytorch.datasets.radfield3d import RadField3DDataset
from RadField3D.pytorch.datasets.radfield3d import RadField3DVoxelwiseDataset
# import the pyTorch compatible datatypes
from RadField3D.pytorch import DataLoaderBuilder
from RadField3D.pytorch.types import DirectionalInput, PositionalInput, TrainingInputData, RadiationField


builder = DataLoaderBuilder(
    "./test_dataset_folder/",
    train_ratio=0.7,
    val_ratio=0.15,
    test_ratio=0.15,
    dataset_class=RadField3DDataset
)

train_dl = builder.build_train_dataloader(
    batch_size=8,
    shuffle=True,
    worker_count=4
)

# iterate over the dataset using fully useable pyTorch classes
for train_data in train_dl:
    input: DirectionalInput | PositionalInput = train_data.input
    field: RadiationField = train_data.ground_truth

TrainingInputData consists of two components metadata (as DirectionalInput or PositionalInput) contains the following information

radiation direction (x, y, z)
radiation origin (x, y, z)
field shape (Cone, Rectangle, Ellipsis)
field shape parameters (opening angle, size at origin, ...)
x-ray tube output spectrum

field (as RadiationField) contains the following information

direct x-ray beam component (as RadiationFieldChannel)
- spectrum per voxel
- fluence per voxel
- statistical error per voxel
scatter field component (as RadiationFieldChannel)
- spectrum per voxel
- fluence per voxel
- statistical error per voxel
geometry (binary density map)

Tracing paths in Cartesian Coordinate Systems

In order to integrate RadFiled3D with other simulation frameworks or applications, one can either take the final results and write it voxel-wise to RadFiled3D or one can already use RadFiled3D during the particle tracking. Therefore, this library offers GridTracers. Each of them implements a different line-segment tracing algorithm to find consecutive voxels that are intersected.

The following GridTracers exists:

SamplingGridTracer: Traces a line between two points in the grid using a sampling approach. In this approach the minimum sampling size is the length of the line segment. If the line segment is longer than the minimum sampling size, which is half the L2-Norm of the voxel size, the line is divided into segments of the minimum sampling size. This approach counts the hits if the line segment is incident to a voxel, only!
BresenhamGridTracer: Traces a line between two points in the grid using the Bresenham algorithm. This algorithm is a line rasterization algorithm that is used to trace a line between two points in a grid. The starting point is excluded as this can only exit a voxel.
LinetracingGridTracer: This class traces a line between two points in the grid using a combination of the SamplingGridTracer and a line tracing algorithm. First the lossy sampling tracer is used to trace the line. Then all adjacent voxels to the voxels that were hit are tested using a line-segment intersection test algorithm.

All those tracers can be created by calling the GridTracerFactory.construct(..) method. The tracers share one single interface method:

def trace(self, p1: vec3, p2: vec3) -> list[int]:

This method takes two points as the definition of the considered line-segment and returns the flat indices of all voxels intersected, that are inside the grid.

Example usage:

from RadFiled3D.RadFiled3D import vec3, GridTracerFactory, GridTracerAlgorithm, CartesianRadiationField, DType

field = CartesianRadiationField(vec3(1.0, 1.0, 1.0), vec3(0.01, 0.01, 0.01))
field.add_channel("test").add_layer("flux", "counts", DType.INT32)
hits_counts = field.get_channel("test").get_layer_as_ndarray("flux")
hits_counts = hits_counts.flatten()

tracer = GridTracerFactory.construct(field, GridTracerAlgorithm.SAMPLING)

indices = tracer.trace(vec3(0.5, 0.5, 0.0), vec3(0.5, 0.85, 1.0))
hits_counts[indices] += 1
grid_shape = field.get_voxel_counts()
hits_counts.reshape((grid_shape.x, grid_shape.y, grid_shape.z))

Faster loading of field series

As the RadFiled3D format possesses a dynamic structure, the loading of a radiation field requires the discovery of channels and layers as well as calculating the binary entry points of channels, layers and voxels. When loading datasets for machine learning, the structure of the fields loaded will likely be constant for each dataset. Therefore, the binary entry points can be precalculated to access only those parts of the RadFiled3D files that are really needed to increase the loading speed and to reduce the needed memory. This is relealized by the FieldAccessors objects.

from RadFiled3D.RadFiled3D import CartesianFieldAccessor, FieldType, uvec3
from RadFiled3D.utils import FieldStore
from RadFiled3D.metadata.v1 import Metadata

accessor: CartesianFieldAccessor = FieldStore.construct_field_accessor("a_file.rf3")
field_type = accessor.get_field_type()
assert field_type == FieldType.CARTESIAN

print(accessor)
field = accessor.access_field("a_similar_file.rf3")
layer = accessor.access_layer("a_similar_file.rf3", "channel1", "layer1")
voxel = accessor.access_voxel("a_similar_file.rf3", "channel1", "layer1", uvec3(0, 0, 0))

FieldAccessors are implemented for the two currently supported coordinate systems: CartesianFieldAccessor and PolarFieldAccessor. Depending on the actual field type, FieldStore.construct_field_accessor(AFile) returns one of them. The pyTorch Datasets are implemented using the FieldAccessor objects to allow for quicker access of datasets. The tests shall act as example code see test_field_accessor.py.

From C++

Simple example on how to create and store a radiation field. Find more in the example file: Example

#include <RadFiled3D/storage/RadiationFieldStore.hpp>
#include <RadFiled3D/RadiationField.hpp>

using namespace RadFiled3D;
using namespace RadFiled3D::Storage;

void main() {
    auto field = std::make_shared<CartesianRadiationField>(glm::vec3(2.5f), glm::vec3(0.05f)); // field extents: 2.5 m x 2.5 m x 2.5 m and voxel extents: 5 cm x 5 cm x 5 cm

    auto metadata = std::make_shared<RadFiled3D::Storage::V1::RadiationFieldMetadata>(
        // learn about the existing data fields from the example file in ./examples/cxx/examples01.cpp
    )

    FieldStore::store(field, metadata, "test_field.rf3", StoreVersion::V1);

    auto field2 = FieldStore::load("test_field.rf3");
}

Available Voxel Datatypes

In general, a C++ Scalar- or HistogramVoxel (and thus layers) can hold any datatype. But in order to deserialize them from a file or use them from Python, there is only a specific list implemented. The Available datatypes are:

C++ Type	RadFiled3D.DType
float	DType.FLOAT32
double	DType.FLOAT64
int	DType.INT32
uint8_t	DType.BYTE
unsigned char	DType.BYTE
char	DType.SCHAR
uint32_t	DType.UINT32
uint64_t	DType.UINT64
unsigned long long	DType.UINT64
glm::vec2	DType.VEC2
glm::vec3	DType.VEC3
glm::vec4	DType.VEC4
HistogramVoxel	DType.HISTOGRAM
AngularResolvedVoxel	DType.ANGULAR

Field Structure

RadFiled3D defines a field structure, that provides the user with the possibility to first define in which kind of space he wants to operate. Therefore one can choose between CartesianRadiationField and PolarRadiationField.

CartesianRadiationField: Segments a room defined by an extent of the room itself and each cuboid voxel into a set of voxels. Each voxel can be addressed by a 3D position (coordinate: x, y, z), a 3D index (number of the voxel in each dimension) or a flat 1D index.
PolarRadiationField: Segements the surface of a unit sphere into surface segments. Each segment (voxel) can be addressed by a 2D position (coordinate: theta, phi), a 2D index (number of the segment in each dimension) or a flat 1D index.

Fields are then partitioned into channels (VoxelGridBuffer/PolarSegmentsBuffer). All channels share the same size and resolution. A channel is again partitioned into layers (VoxelGrid/PolarSegment). Each layer holds the actual voxel data and can be constructed from various data types (float, double, uint32_t, uint64_t, glm::vec2, glm::vec3, glm::vec4, N-D-Histogram (list of floats)). Additionally, a layer has a unit string assigned to it as well as a statistical uncertainty to perserve those information.

Dependencies

RadFiled3D comes with a possibly low amount of dependencies. We integrated the OpenGL Math Library (GLM) just to provide those datatypes out of the box and as GLM is a head-only library we suspect no issues by doing so.

All C++ dependencies (Will be fetched by CMake):

All python dependencies:

PyBind11
rich
numpy
Optional:
- pyTorch

Download files

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

Source Distribution

radfiled3d-1.3.6.tar.gz (135.2 kB view details)

Uploaded Aug 3, 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.

radfiled3d-1.3.6-cp314-cp314-win_amd64.whl (1.4 MB view details)

Uploaded Aug 3, 2026 CPython 3.14Windows x86-64

radfiled3d-1.3.6-cp314-cp314-musllinux_1_2_x86_64.whl (1.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.14musllinux: musl 1.2+ x86-64

radfiled3d-1.3.6-cp314-cp314-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (808.4 kB view details)

Uploaded Aug 3, 2026 CPython 3.14manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

radfiled3d-1.3.6-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (837.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.14manylinux: glibc 2.17+ x86-64

radfiled3d-1.3.6-cp313-cp313-win_amd64.whl (1.3 MB view details)

Uploaded Aug 3, 2026 CPython 3.13Windows x86-64

radfiled3d-1.3.6-cp313-cp313-musllinux_1_2_x86_64.whl (1.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.13musllinux: musl 1.2+ x86-64

radfiled3d-1.3.6-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (808.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.13manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

radfiled3d-1.3.6-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (836.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.13manylinux: glibc 2.17+ x86-64

radfiled3d-1.3.6-cp312-cp312-win_amd64.whl (1.3 MB view details)

Uploaded Aug 3, 2026 CPython 3.12Windows x86-64

radfiled3d-1.3.6-cp312-cp312-musllinux_1_2_x86_64.whl (1.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.12musllinux: musl 1.2+ x86-64

radfiled3d-1.3.6-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (807.4 kB view details)

Uploaded Aug 3, 2026 CPython 3.12manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

radfiled3d-1.3.6-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (836.1 kB view details)

Uploaded Aug 3, 2026 CPython 3.12manylinux: glibc 2.17+ x86-64

radfiled3d-1.3.6-cp311-cp311-musllinux_1_2_x86_64.whl (1.8 MB view details)

Uploaded Aug 3, 2026 CPython 3.11musllinux: musl 1.2+ x86-64

radfiled3d-1.3.6-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (804.3 kB view details)

Uploaded Aug 3, 2026 CPython 3.11manylinux: glibc 2.27+ x86-64manylinux: glibc 2.28+ x86-64

radfiled3d-1.3.6-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (831.0 kB view details)

Uploaded Aug 3, 2026 CPython 3.11manylinux: glibc 2.17+ x86-64

File details

Details for the file radfiled3d-1.3.6.tar.gz.

File metadata

Download URL: radfiled3d-1.3.6.tar.gz
Upload date: Aug 3, 2026
Size: 135.2 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for radfiled3d-1.3.6.tar.gz
Algorithm	Hash digest
SHA256	`057c70daddbfa3153de92e4580280c74fdbbc56a52b2a880b5f532130eb5610d`
MD5	`7a9944451506b59ae7d013503f7da6cf`
BLAKE2b-256	`869981804fbeb90a154cd95b614b7f8b26d60f5a3db0c9d5bc3257afb169f2f9`

Algorithm	Hash digest
SHA256	`73b306630eb1f64408cb7a6a6a3fdeaaf02e9b7b14ebe77affd0c97e038260c8`
MD5	`1c6555fa66812a4cb6bfc355ca8b8af9`
BLAKE2b-256	`ffc6ba008beda4a6a98af6d87abf8482442e0c9ab1b95bec6dc327d82500336c`

Algorithm	Hash digest
SHA256	`7ba866b52e80f95d76561c70bd6d0eb4f3516f8f46998b081df84e52fa6b6282`
MD5	`049c624c03f21636ca76711d2c1d6a4f`
BLAKE2b-256	`559073ebe42ba246f873aa3a18a3352990941cd11e82c38b6f8ca0c35146ebca`

Algorithm	Hash digest
SHA256	`9b9e9e3bc5f703f38669655c649ff518891f7a6baebc2c44b3ce9e996698d15b`
MD5	`c79ccd8534719a6c1ead4c63bf5cc260`
BLAKE2b-256	`7fc01784573803ef7617d0a273e1f0b2ca2f7958c78e404e70ec46b76708f20c`

Algorithm	Hash digest
SHA256	`0ff5b4ca3bced23d29122cc9eca561706056e072c94d716f40f2aa0345d70d88`
MD5	`d3450b589d959093030044bf849db5e5`
BLAKE2b-256	`96b13259dfeef9fc43831a8148292b4812feb3aed63385d84ee725cacced4d89`

Algorithm	Hash digest
SHA256	`12c5e4f9a8961f64f1f5ea574d929eb6d96bccbf2e3d6b12586c4649cd341371`
MD5	`51f69b2a02878d8b07a69a22885c40ee`
BLAKE2b-256	`762169b21f5d28effd8598126f2c5a96ebc4bad105b6670ca2c5b917e41a8735`

Algorithm	Hash digest
SHA256	`273ee19bd2321be4cac961f1cb87931b49b4e772e850d32b5e7aeef44a646ba2`
MD5	`56a38cd5379aa0223b5f439c8429cf77`
BLAKE2b-256	`cf2e571ff439e19216820771cfa8df313845a469409ff4ee4787fb9471677805`

Algorithm	Hash digest
SHA256	`36b3dd8ab74eeb99fb76b856c96d5daaff3d5063575fdc8ab0c4f18f949f8ac3`
MD5	`eeea0bfb82690c984390cf793b701c38`
BLAKE2b-256	`3dfe70388e574f4ae44c85240d5742981f307c705d5945db16fab1d5d6091e65`

Algorithm	Hash digest
SHA256	`0d99ed9e92b9d9ba728994ee35964492c91140b3f86ce7f8a52f9d5ec99ee098`
MD5	`068e547d25769ece971fe245229f6697`
BLAKE2b-256	`f6d92ab2b42a56fe34fadd5bd77b5c3eefbb2f25c300f4f4365a0656f37159c1`

Algorithm	Hash digest
SHA256	`7c2837e9b0818b82b19ae4e83a3e2e4bdeaafab6b7a693a6c0f8261d747be68d`
MD5	`323cda1b017199eb8900b1d0e7ebc41a`
BLAKE2b-256	`cf0ff9ae3b96fde7713a96e2d5a09e7d89b90e40d7ffda2fecfd7ad986b52c27`

Algorithm	Hash digest
SHA256	`2e911199012e2d645b2d1fd22a1a78ef248dd8fb5c811281f652183482752aa4`
MD5	`a5b3f1e7f6af0b9ae613476c12e746ff`
BLAKE2b-256	`b35759808951a9592f0d9d207371b6be480e2011924ce38fa16c552bbedbb599`

Algorithm	Hash digest
SHA256	`5b5f3505f531a6ad1391f9775e0000f7967f98c58c413104512f64b9e17f39d6`
MD5	`ea9f2b2d89f4fa73e3d3c3cdf0292e9a`
BLAKE2b-256	`3321c8eb2cc459dbcb60f80df6beb5a007753235230a5b0958a1f8b6cd047b15`

Algorithm	Hash digest
SHA256	`bdcf63d4c1895e40e34f772d40d29d06baee97b255db38a5299fca2aaac20d1a`
MD5	`0bbb7c36eea023664b1f7942d5807ad3`
BLAKE2b-256	`491308d703633acd8ccc805d896fc6ae3c7c48b887effb7e70c7a09e7e6b670b`

Algorithm	Hash digest
SHA256	`43332467d12176a650a4efe267333b5f9f8ff0dd4b2f491a2018a3eaeffd2196`
MD5	`f5466af17f2e914d8664fd336cb686cd`
BLAKE2b-256	`f95e410c4ca5d50f027cf306f63dea88fb9693a929858d794e1e5d18f342b368`

Algorithm	Hash digest
SHA256	`c01101c128d5a75e6fa42eae7701b64fb4ddff36ce3fa8ceddf859f3265cb9c0`
MD5	`3424d3d8a13805fde20e5dd1c8f7da33`
BLAKE2b-256	`c9fe0503bf9dbfc8e3642d7147818e02aa8b83abc6ed7132a08dc0cc4332a701`

Algorithm	Hash digest
SHA256	`43dc66f3e2379cc923a0f3c7ce0183dfacb33e05fcc0f8da31cc91d31fd23f62`
MD5	`57a300607d1cc89116115b5990e1489a`
BLAKE2b-256	`155757efc252930263e745607f5c713261281fa31bf8f85e292b692ee7faf155`

RadFiled3D 1.3.6

RadFiled3D

🌟 Why Use RadFiled3D

Table of Contents

Building and Installing

Installing from PyPi

Installing from Source

Prerequisites

CMake

Python

Installing locally

Building a wheel

Getting Started

From Python

Integrating with pyTorch

Direct integration with RadField3D datasets

Tracing paths in Cartesian Coordinate Systems

Faster loading of field series

From C++

Available Voxel Datatypes

Field Structure

Dependencies

Project links

Key dates

1 maintainer

Credits

License

Requires

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

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

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

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance