Class that inherits from both `torch.Tensor` and `MetaObj`, adding support for metadata. Metadata is stored in the form of a dictionary. Nes

``Compose`` provides the ability to chain a series of callables together in a sequential manner. Each transform in the sequence must take a s

Constructs a convolution with normalization, optional dropout, and optional activation layers:: -- (Conv|ConvTrans) -- (Norm -- Dropout

Dictionary-based wrapper of :py:class:`monai.transforms.LoadImage`, It can load both image data and metadata. When loading a list of files in

Provides an iterable over the given `dataset`. It inherits the PyTorch DataLoader and adds enhanced `collate_fn` and `worker_fn` by default.

Load image file or files from provided path based on reader. If reader is not specified, this class automatically chooses readers based o

The primary configuration parser. It traverses a structured config (in the form of nested Python dict or list), creates ``ConfigItem``, and a

Dictionary-based wrapper of :py:class:`monai.transforms.EnsureChannelFirst`.

Unet network with timestep embedding and attention mechanisms for conditioning based on Rombach et al. "High-Resolution Image Synthesis with

A generic dataset with a length property and an optional callable data transform when fetching a data sample. If passing slicing indices,

Denoising diffusion probabilistic models (DDPMs) explores the connections between denoising score matching and Langevin dynamics sampling. Ba

Splits the input into patches with sliding window strategy and a possible overlap. It also allows offsetting the starting position and filter

Computes the Normalized Surface Dice (NSD) for each batch sample and class of predicted segmentations `y_pred` and corresponding reference se

Autoencoder model with KL-regularized latent space based on Rombach et al. "High-Resolution Image Synthesis with Latent Diffusion Models" htt

Vector-Quantised Variational Autoencoder (VQ-VAE) used in Morphology-preserving Autoregressive 3D Generative Modelling of the Brain by Tudosi

UNet network with timestep embedding and attention mechanisms for conditioning, with added SPADE normalization for semantic conditioning (Par

FocalLoss is an extension of BCEWithLogitsLoss that down-weights loss from high confidence correct predictions. Reimplementation of the

Adjust or add the channel dimension of input data to ensure `channel_first` shape. This extracts the `original_channel_dim` info from provid

Dictionary-based wrapper of :py:class:`monai.transforms.Spacing`. This transform assumes the ``data`` dictionary has a key for the input

A decorator and context manager for NVIDIA Tools Extension (NVTX) Range for profiling. When used as a decorator it encloses a specific method

Converts the input image to a tensor without applying any other transformations. Input data can be PyTorch Tensor, numpy array, list, diction

Compute average Dice loss between two tensors. It can support both multi-classes and multi-labels tasks. The data `input` (BNHW[D] where N is

Autoencoder model with KL-regularized latent space based on Rombach et al. "High-Resolution Image Synthesis with Latent Diffusion Models" htt

Dictionary-based wrapper of :py:class:`monai.transforms.ScaleIntensity`. Scale the intensity of input image to the given value range (minv, m

Enhanced version of UNet which has residual units implemented with the ResidualUnit class. The residual part uses a convolution to change the

Sliding window method for model inference, with `sw_batch_size` windows for every model.forward(). Usage example can be found in the :py:

``SomeOf`` samples a different sequence of transforms to apply each time it is called. It can be configured to sample a fixed or varying num

Reverses the order of elements along the given spatial axis. Preserves shape. See `torch.flip` documentation for additional details: http

Applies a convolution filter to the input image. Args: filter: A string specifying the filter, a custom filter as ``torc

Utility transform to invert the previously applied transforms. Taking the ``transform`` previously applied on ``orig_keys``, this ``Invertd`

``OneOf`` provides the ability to randomly choose one transform out of a list of callables with pre-defined probabilities for each. Args

A rectified flow scheduler for guiding the diffusion process in a generative model. Supports uniform and logit-normal sampling methods, time

A self-attention block, based on: "Dosovitskiy et al., An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale <https://ar

Specification for the config-based bundle workflow. Standardized the `initialize`, `run`, `finalize` behavior in a config-based training, eva

Control network for diffusion models based on Zhang and Agrawala "Adding Conditional Control to Text-to-Image Diffusion Models" (https://arxi

See also: - :py:class:`monai.losses.dice.DiceLoss` - :py:class:`monai.losses.dice.GeneralizedDiceLoss` - :py:class:`monai

Dataset with cache mechanism that can load data and cache deterministic transforms' result during training. By caching the results of non-ra

Apply clip based on the intensity distribution of input image. If `sharpness_factor` is provided, the intensity values will be soft clipped a

U-Net network with timestep embedding and attention mechanisms for conditioning based on Rombach et al. "High-Resolution Image Synthesis with

Load medical images based on ITK library. All the supported image formats can be found at: https://github.com/InsightSoftwareConsortium/I

Convert the input tensor/array into discrete values, possible operations are: - `argmax`. - threshold input value to binary va

Dictionary-based wrapper of :py:class:`monai.transforms.ClipIntensityPercentiles`. Clip the intensity values of input image to a specific ran

Residual block with timestep conditioning. Args: spatial_dims: The number of spatial dimensions. in_channels: number of inpu

Dictionary-based wrapper of :py:class:`monai.transforms.Lambda`. For example: .. code-block:: python :emphasize-lines: 2

Rotate an array by 90 degrees in the plane specified by `axes`. See `torch.rot90` for additional details: https://pytorch.org/docs/stable

A cross-attention block, based on: "Dosovitskiy et al., An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale <https://a

Resize the input image to given spatial size (with scaling, not cropping/padding). Implemented using :py:class:`torch.nn.functional.interpola

Compute confidence map from an ultrasound image. This transform uses the method introduced by Karamalis et al. in https://doi.org/10.1016/j.media.

DiffusionInferer takes a trained diffusion model and a scheduler and can be used to perform a signal forward pass for a training iteration, a

Dictionary-based wrapper of :py:class:`monai.transforms.Orientation`. This transform assumes the channel-first input format. In the case

Persistent storage of pre-computed values to efficiently manage larger than memory dictionary format data, it can operate transforms for spec

General purpose cropper to produce sub-volume region of interest (ROI). If a dimension of the expected ROI size is larger than the input imag

Dictionary-based wrapper of :py:class:`monai.transforms.SpatialCrop`. General purpose cropper to produce sub-volume region of interest (ROI).

Activation operations, typically `Sigmoid` or `Softmax`. Args: sigmoid: whether to execute sigmoid function on model output before t

Perform spatial self-attention on the input tensor. The input tensor is reshaped to B x (x_dim * y_dim [ * z_dim]) x C, where C is the number of

Dictionary-based wrapper of :py:class:`monai.transforms.EnsureType`. Ensure the input data to be a PyTorch Tensor or numpy array, support: `

A patch embedding block, based on: "Dosovitskiy et al., An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale <https://a

Dictionary-based version :py:class:`monai.transforms.RandRotate90`. With probability `prob`, input arrays are rotated by 90 degrees in th

Dictionary-based wrapper of :py:class:`monai.transforms.SimulateDelay`.

Resample input image into the specified `pixdim`. This transform is capable of lazy execution. See the :ref:`Lazy Resampling topic<lazy_resa

upsampling, concatenation with the encoder feature map, two convolutions

Read whole slide images and extract patches using different backend libraries Args: backend: the name of backend whole slide image r

CheckpointSaver acts as an Ignite handler to save checkpoint data into files. It supports to save according to metrics result, epoch number,

Compute the generalised Dice loss defined in: Sudre, C. et. al. (2017) Generalised Dice overlap as a deep learning loss function

Generate per-landmark Gaussian heatmaps for 2D or 3D coordinates. Notes: - Coordinates are interpreted in voxel units and expected i

Dictionary-based wrapper of :py:class:`monai.transforms.Resize`. This transform is capable of lazy execution. See the :ref:`Lazy Resampling

Apply statistical operation to a sample (image/ndarray/tensor). Notes: Percentile operation uses a partial function that embeds diff

Scale the intensity of input image to the given value range (minv, maxv). If `minv` and `maxv` not provided, use `factor` to scale image by `

Standard supervised evaluation method with image and label(optional), inherits from evaluator and Workflow. Args: device: an object

Standard supervised training method with image and label, inherits from ``Trainer`` and ``Workflow``. Args: device: an object repres

Converts the input data to numpy array, can support list or tuple of numbers and PyTorch Tensor. Args: dtype: target data type when

AutoencoderKL with custom MaisiEncoder and MaisiDecoder. Args: spatial_dims: Number of spatial dimensions (1D, 2D, 3D). in_c

An extension of the PersistentDataset using direct memory access(DMA) data path between GPU memory and storage, thus avoiding a bounce buffer

Load NPY or NPZ format data based on Numpy library, they can be arrays or pickled objects. A typical usage is to load the `mask` data for cla

Dictionary-based version :py:class:`monai.transforms.RandFlip`. See `numpy.flip` for additional details. https://docs.scipy.org/doc/nump

StatsHandler defines a set of Ignite Event-handlers for all the log printing logics. It can be used for any Ignite Engine(trainer, validator

Dictionary-based wrapper of :py:class:`monai.transforms.ToTensor`.

A transformer block, based on: "Dosovitskiy et al., An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale <https://arxiv

A CNN module that can be used for UNETR, based on: "Hatamizadeh et al., UNETR: Transformers for 3D Medical Image Segmentation <https://arxiv.

Compute the Calibration Error between predicted probabilities and ground truth labels. **Why Calibration Matters:** A well-calibrated c

Copy specified items from data dictionary and save with different key names. It can copy several items together and copy several times.

Denoising diffusion implicit models is a scheduler that extends the denoising procedure introduced in denoising diffusion probabilistic model

Decoder-only (Autoregressive) Transformer model. Args: num_tokens: Number of tokens in the vocabulary. max_seq_len: Maximum seque

Apply Gaussian smooth to the input data based on specified `sigma` parameter. A default value `sigma=1.0` is provided for reference. Arg

Apply a user-defined lambda as a transform. For example: .. code-block:: python :emphasize-lines: 2 image = np.ones((1

Change the input image's orientation into the specified based on `axcodes`. This transform is capable of lazy execution. See the :ref:`Lazy

Randomly changes image intensity with gamma transform. Each pixel/voxel intensity is updated as: x = ((x - min) / intensity_range) ^ gam

Dictionary-based wrapper of :py:class:`monai.transforms.RandAffine`. This transform is capable of lazy execution. See the :ref:`Lazy Resampl

Randomly flips the image along axes. Preserves shape. See numpy.flip for additional details. https://docs.scipy.org/doc/numpy/reference/g