Hugging Face's logo

Transformers documentation

MaskFormer

You are viewing v4.20.1 version. A newer version v4.46.3 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

MaskFormer

This is a recently introduced model so the API hasn’t been tested extensively. There may be some bugs or slight breaking changes to fix it in the future. If you see something strange, file a Github Issue.

Overview

The MaskFormer model was proposed in Per-Pixel Classification is Not All You Need for Semantic Segmentation by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov. MaskFormer addresses semantic segmentation with a mask classification paradigm instead of performing classic pixel-level classification.

The abstract from the paper is the following:

Modern approaches typically formulate semantic segmentation as a per-pixel classification task, while instance-level segmentation is handled with an alternative mask classification. Our key insight: mask classification is sufficiently general to solve both semantic- and instance-level segmentation tasks in a unified manner using the exact same model, loss, and training procedure. Following this observation, we propose MaskFormer, a simple mask classification model which predicts a set of binary masks, each associated with a single global class label prediction. Overall, the proposed mask classification-based method simplifies the landscape of effective approaches to semantic and panoptic segmentation tasks and shows excellent empirical results. In particular, we observe that MaskFormer outperforms per-pixel classification baselines when the number of classes is large. Our mask classification-based method outperforms both current state-of-the-art semantic (55.6 mIoU on ADE20K) and panoptic segmentation (52.7 PQ on COCO) models.

Tips:

  • MaskFormer’s Transformer decoder is identical to the decoder of DETR. During training, the authors of DETR did find it helpful to use auxiliary losses in the decoder, especially to help the model output the correct number of objects of each class. If you set the parameter use_auxilary_loss of MaskFormerConfig to True, then prediction feedforward neural networks and Hungarian losses are added after each decoder layer (with the FFNs sharing parameters).
  • If you want to train the model in a distributed environment across multiple nodes, then one should update the get_num_masks function inside in the MaskFormerLoss class of modeling_maskformer.py. When training on multiple nodes, this should be set to the average number of target masks across all nodes, as can be seen in the original implementation here.
  • One can use MaskFormerFeatureExtractor to prepare images for the model and optional targets for the model.
  • To get the final segmentation, depending on the task, you can call post_process_semantic_segmentation() or post_process_panoptic_segmentation(). Both tasks can be solved using MaskFormerForInstanceSegmentation output, the latter needs an additional is_thing_map to know which instances must be merged together..

The figure below illustrates the architecture of MaskFormer. Taken from the original paper.

This model was contributed by francesco. The original code can be found here.

MaskFormer specific outputs

class transformers.models.maskformer.modeling_maskformer.MaskFormerModelOutput

< >

( encoder_last_hidden_state: typing.Optional[torch.FloatTensor] = None pixel_decoder_last_hidden_state: typing.Optional[torch.FloatTensor] = None transformer_decoder_last_hidden_state: typing.Optional[torch.FloatTensor] = None encoder_hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None pixel_decoder_hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None transformer_decoder_hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None attentions: typing.Optional[typing.Tuple[torch.FloatTensor]] = None )

Parameters

  • encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the encoder model (backbone).
  • pixel_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the pixel decoder model (FPN).
  • transformer_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Last hidden states (final feature map) of the last stage of the transformer decoder model.
  • encoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the encoder model at the output of each stage.
  • pixel_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the pixel decoder model at the output of each stage.
  • transformer_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, sequence_length, hidden_size). Hidden-states (also called feature maps) of the transformer decoder at the output of each stage.
  • hidden_states tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor containing encoder_hidden_states, pixel_decoder_hidden_states and decoder_hidden_states
  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length). Attentions weights from Detr’s decoder after the attention softmax, used to compute the weighted average in the self-attention heads.

Class for outputs of MaskFormerModel. This class returns all the needed hidden states to compute the logits.

class transformers.models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentationOutput

< >

( loss: typing.Optional[torch.FloatTensor] = None class_queries_logits: FloatTensor = None masks_queries_logits: FloatTensor = None auxiliary_logits: FloatTensor = None encoder_last_hidden_state: typing.Optional[torch.FloatTensor] = None pixel_decoder_last_hidden_state: typing.Optional[torch.FloatTensor] = None transformer_decoder_last_hidden_state: typing.Optional[torch.FloatTensor] = None encoder_hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None pixel_decoder_hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None transformer_decoder_hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None hidden_states: typing.Optional[typing.Tuple[torch.FloatTensor]] = None attentions: typing.Optional[typing.Tuple[torch.FloatTensor]] = None )

Parameters

  • loss (torch.Tensor, optional) — The computed loss, returned when labels are present.
  • class_queries_logits (torch.FloatTensor) — A tensor of shape (batch_size, num_queries, height, width) representing the proposed masks for each query.
  • masks_queries_logits (torch.FloatTensor) — A tensor of shape (batch_size, num_queries, num_labels + 1) representing the proposed classes for each query. Note the + 1 is needed because we incorporate the null class.
  • encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the encoder model (backbone).
  • pixel_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the pixel decoder model (FPN).
  • transformer_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Last hidden states (final feature map) of the last stage of the transformer decoder model.
  • encoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the encoder model at the output of each stage.
  • pixel_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the pixel decoder model at the output of each stage.
  • transformer_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, sequence_length, hidden_size). Hidden-states of the transformer decoder at the output of each stage.
  • hidden_states tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor containing encoder_hidden_states, pixel_decoder_hidden_states and decoder_hidden_states.
  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length). Attentions weights from Detr’s decoder after the attention softmax, used to compute the weighted average in the self-attention heads.

Class for outputs of MaskFormerForInstanceSegmentation.

This output can be directly passed to post_process_segmentation() or post_process_panoptic_segmentation() depending on the task. Please, see [`~MaskFormerFeatureExtractor] for details regarding usage.

MaskFormerConfig

class transformers.MaskFormerConfig

< >

( fpn_feature_size: int = 256 mask_feature_size: int = 256 no_object_weight: float = 0.1 use_auxiliary_loss: bool = False backbone_config: typing.Optional[typing.Dict] = None decoder_config: typing.Optional[typing.Dict] = None init_std: float = 0.02 init_xavier_std: float = 1.0 dice_weight: float = 1.0 cross_entropy_weight: float = 1.0 mask_weight: float = 20.0 output_auxiliary_logits: typing.Optional[bool] = None **kwargs )

Parameters

  • mask_feature_size (int, optional, defaults to 256) — The masks’ features size, this value will also be used to specify the Feature Pyramid Network features’ size.
  • no_object_weight (float, optional, defaults to 0.1) — Weight to apply to the null (no object) class.
  • use_auxiliary_loss(bool, optional, defaults to False) — If True MaskFormerForInstanceSegmentationOutput will contain the auxiliary losses computed using the logits from each decoder’s stage.
  • backbone_config (Dict, optional) — The configuration passed to the backbone, if unset, the configuration corresponding to swin-base-patch4-window12-384 will be used.
  • decoder_config (Dict, optional) — The configuration passed to the transformer decoder model, if unset the base config for detr-resnet-50 will be used.
  • init_std (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
  • init_xavier_std (float, optional, defaults to 1) — The scaling factor used for the Xavier initialization gain in the HM Attention map module.
  • dice_weight (float, optional, defaults to 1.0) — The weight for the dice loss.
  • cross_entropy_weight (float, optional, defaults to 1.0) — The weight for the cross entropy loss.
  • mask_weight (float, optional, defaults to 20.0) — The weight for the mask loss.
  • output_auxiliary_logits (bool, optional) — Should the model output its auxiliary_logits or not.

This is the configuration class to store the configuration of a MaskFormerModel. It is used to instantiate a MaskFormer model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the MaskFormer facebook/maskformer-swin-base-ade architecture trained on ADE20k-150.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Currently, MaskFormer only supports the Swin Transformer as backbone.

Examples:

>>> from transformers import MaskFormerConfig, MaskFormerModel

>>> # Initializing a MaskFormer facebook/maskformer-swin-base-ade configuration
>>> configuration = MaskFormerConfig()

>>> # Initializing a model from the facebook/maskformer-swin-base-ade style configuration
>>> model = MaskFormerModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config

from_backbone_and_decoder_configs

< >

( backbone_config: PretrainedConfig decoder_config: PretrainedConfig **kwargs ) → MaskFormerConfig

Parameters

Returns

MaskFormerConfig

An instance of a configuration object

Instantiate a MaskFormerConfig (or a derived class) from a pre-trained backbone model configuration and DETR model configuration.

to_dict

< >

( ) → Dict[str, any]

Returns

Dict[str, any]

Dictionary of all the attributes that make up this configuration instance,

Serializes this instance to a Python dictionary. Override the default to_dict().

MaskFormerFeatureExtractor

class transformers.MaskFormerFeatureExtractor

< >

( do_resize = True size = 800 max_size = 1333 resample = <Resampling.BILINEAR: 2> size_divisibility = 32 do_normalize = True image_mean = None image_std = None ignore_index = None reduce_labels = False **kwargs )

Parameters

  • do_resize (bool, optional, defaults to True) — Whether to resize the input to a certain size.
  • size (int, optional, defaults to 800) — Resize the input to the given size. Only has an effect if do_resize is set to True. If size is a sequence like (width, height), output size will be matched to this. If size is an int, smaller edge of the image will be matched to this number. i.e, if height > width, then image will be rescaled to (size * height / width, size).
  • max_size (int, optional, defaults to 1333) — The largest size an image dimension can have (otherwise it’s capped). Only has an effect if do_resize is set to True.
  • resample (int, optional, defaults to PIL.Image.BILINEAR) — An optional resampling filter. This can be one of PIL.Image.NEAREST, PIL.Image.BOX, PIL.Image.BILINEAR, PIL.Image.HAMMING, PIL.Image.BICUBIC or PIL.Image.LANCZOS. Only has an effect if do_resize is set to True.
  • size_divisibility (int, optional, defaults to 32) — Some backbones need images divisible by a certain number. If not passed, it defaults to the value used in Swin Transformer.
  • do_normalize (bool, optional, defaults to True) — Whether or not to normalize the input with mean and standard deviation.
  • image_mean (int, optional, defaults to [0.485, 0.456, 0.406]) — The sequence of means for each channel, to be used when normalizing images. Defaults to the ImageNet mean.
  • image_std (int, optional, defaults to [0.229, 0.224, 0.225]) — The sequence of standard deviations for each channel, to be used when normalizing images. Defaults to the ImageNet std.
  • ignore_index (int, optional) — Value of the index (label) to be removed from the segmentation maps.
  • reduce_labels (bool, optional, defaults to False) — Whether or not to reduce all label values of segmentation maps by 1. Usually used for datasets where 0 is used for background, and background itself is not included in all classes of a dataset (e.g. ADE20k). The background label will be replaced by ignore_index.

Constructs a MaskFormer feature extractor. The feature extractor can be used to prepare image(s) and optional targets for the model.

This feature extractor inherits from FeatureExtractionMixin which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.

__call__

< >

( images: typing.Union[PIL.Image.Image, numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[PIL.Image.Image], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]] segmentation_maps: typing.Union[PIL.Image.Image, numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[PIL.Image.Image], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]] = None pad_and_return_pixel_mask: typing.Optional[bool] = True instance_id_to_semantic_id: typing.Union[typing.Dict[int, int], NoneType] = None return_tensors: typing.Union[str, transformers.utils.generic.TensorType, NoneType] = None **kwargs ) → BatchFeature

Parameters

  • images (PIL.Image.Image, np.ndarray, torch.Tensor, List[PIL.Image.Image], List[np.ndarray], List[torch.Tensor]) — The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch tensor. In case of a NumPy array/PyTorch tensor, each image should be of shape (C, H, W), where C is a number of channels, H and W are image height and width.
  • segmentation_maps (PIL.Image.Image, np.ndarray, torch.Tensor, List[PIL.Image.Image], List[np.ndarray], List[torch.Tensor], optional) — Optionally, the corresponding semantic segmentation maps with the pixel-wise annotations.
  • pad_and_return_pixel_mask (bool, optional, defaults to True) — Whether or not to pad images up to the largest image in a batch and create a pixel mask.

    If left to the default, will return a pixel mask that is:

    • 1 for pixels that are real (i.e. not masked),
    • 0 for pixels that are padding (i.e. masked).
  • instance_id_to_semantic_id (Dict[int, int], optional) — If passed, we treat segmentation_maps as an instance segmentation map where each pixel represents an instance id. To convert it to a binary mask of shape (batch, num_labels, height, width) we need a dictionary mapping instance ids to label ids to create a semantic segmentation map.
  • return_tensors (str or TensorType, optional) — If set, will return tensors instead of NumPy arrays. If set to 'pt', return PyTorch torch.Tensor objects.

Returns

BatchFeature

A BatchFeature with the following fields:

  • pixel_values — Pixel values to be fed to a model.
  • pixel_mask — Pixel mask to be fed to a model (when pad_and_return_pixel_mask=True or if “pixel_mask” is in self.model_input_names).
  • mask_labels — Optional list of mask labels of shape (labels, height, width) to be fed to a model (when annotations are provided).
  • class_labels — Optional list of class labels of shape (labels) to be fed to a model (when annotations are provided). They identify the labels of mask_labels, e.g. the label of mask_labels[i][j] if class_labels[i][j].

Main method to prepare for the model one or several image(s) and optional annotations. Images are by default padded up to the largest image in a batch, and a pixel mask is created that indicates which pixels are real/which are padding.

MaskFormer addresses semantic segmentation with a mask classification paradigm, thus input segmentation maps will be converted to lists of binary masks and their respective labels. Let’s see an example, assuming segmentation_maps = [[2,6,7,9]], the output will contain mask_labels = [[1,0,0,0],[0,1,0,0],[0,0,1,0],[0,0,0,1]] (four binary masks) and class_labels = [2,6,7,9], the labels for each mask.

NumPy arrays and PyTorch tensors are converted to PIL images when resizing, so the most efficient is to pass PIL images.

encode_inputs

< >

( pixel_values_list: typing.List[ForwardRef('np.ndarray')] segmentation_maps: typing.Union[PIL.Image.Image, numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[PIL.Image.Image], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]] = None pad_and_return_pixel_mask: bool = True instance_id_to_semantic_id: typing.Union[typing.Dict[int, int], NoneType] = None return_tensors: typing.Union[str, transformers.utils.generic.TensorType, NoneType] = None ) → BatchFeature

Parameters

  • pixel_values_list (List[torch.Tensor]) — List of images (pixel values) to be padded. Each image should be a tensor of shape (channels, height, width).
  • segmentation_maps (PIL.Image.Image, np.ndarray, torch.Tensor, List[PIL.Image.Image], List[np.ndarray], List[torch.Tensor], optional) — The corresponding semantic segmentation maps with the pixel-wise annotations.
  • pad_and_return_pixel_mask (bool, optional, defaults to True) — Whether or not to pad images up to the largest image in a batch and create a pixel mask.

    If left to the default, will return a pixel mask that is:

    • 1 for pixels that are real (i.e. not masked),
    • 0 for pixels that are padding (i.e. masked).
  • instance_id_to_semantic_id (Dict[int, int], optional) — If passed, we treat segmentation_maps as an instance segmentation map where each pixel represents an instance id. To convert it to a binary mask of shape (batch, num_labels, height, width) we need a dictionary mapping instance ids to label ids to create a semantic segmentation map.
  • return_tensors (str or TensorType, optional) — If set, will return tensors instead of NumPy arrays. If set to 'pt', return PyTorch torch.Tensor objects.

Returns

BatchFeature

A BatchFeature with the following fields:

  • pixel_values — Pixel values to be fed to a model.
  • pixel_mask — Pixel mask to be fed to a model (when pad_and_return_pixel_mask=True or if “pixel_mask” is in self.model_input_names).
  • mask_labels — Optional list of mask labels of shape (labels, height, width) to be fed to a model (when annotations are provided).
  • class_labels — Optional list of class labels of shape (labels) to be fed to a model (when annotations are provided). They identify the labels of mask_labels, e.g. the label of mask_labels[i][j] if class_labels[i][j].

Pad images up to the largest image in a batch and create a corresponding pixel_mask.

MaskFormer addresses semantic segmentation with a mask classification paradigm, thus input segmentation maps will be converted to lists of binary masks and their respective labels. Let’s see an example, assuming segmentation_maps = [[2,6,7,9]], the output will contain mask_labels = [[1,0,0,0],[0,1,0,0],[0,0,1,0],[0,0,0,1]] (four binary masks) and class_labels = [2,6,7,9], the labels for each mask.

post_process_segmentation

< >

( outputs: MaskFormerForInstanceSegmentationOutput target_size: typing.Tuple[int, int] = None ) → torch.Tensor

Parameters

  • outputs (MaskFormerForInstanceSegmentationOutput) — The outputs from MaskFormerForInstanceSegmentation.
  • target_size (Tuple[int, int], optional) — If set, the masks_queries_logits will be resized to target_size.

Returns

torch.Tensor

A tensor of shape (batch_size, num_labels, height, width).

Converts the output of MaskFormerForInstanceSegmentationOutput into image segmentation predictions. Only supports PyTorch.

post_process_semantic_segmentation

< >

( outputs: MaskFormerForInstanceSegmentationOutput target_size: typing.Tuple[int, int] = None ) → torch.Tensor

Parameters

Returns

torch.Tensor

A tensor of shape batch_size, height, width.

Converts the output of MaskFormerForInstanceSegmentationOutput into semantic segmentation predictions. Only supports PyTorch.

post_process_panoptic_segmentation

< >

( outputs: MaskFormerForInstanceSegmentationOutput object_mask_threshold: float = 0.8 overlap_mask_area_threshold: float = 0.8 label_ids_to_fuse: typing.Optional[typing.Set[int]] = None ) → List[Dict]

Parameters

  • outputs (MaskFormerForInstanceSegmentationOutput) — The outputs from MaskFormerForInstanceSegmentation.
  • object_mask_threshold (float, optional, defaults to 0.8) — The object mask threshold.
  • overlap_mask_area_threshold (float, optional, defaults to 0.8) — The overlap mask area threshold to use.
  • label_ids_to_fuse (Set[int], optional) — The labels in this state will have all their instances be fused together. For instance we could say there can only be one sky in an image, but several persons, so the label ID for sky would be in that set, but not the one for person.

Returns

List[Dict]

A list of dictionaries, one per image, each dictionary containing two keys:

  • segmentation — a tensor of shape (height, width) where each pixel represents a segment_id.
  • segments — a dictionary with the following keys
    • id — an integer representing the segment_id.
    • label_id — an integer representing the segment’s label.
    • was_fused — a boolean, True if label_id was in label_ids_to_fuse, False otherwise.

Converts the output of MaskFormerForInstanceSegmentationOutput into image panoptic segmentation predictions. Only supports PyTorch.

MaskFormerModel

class transformers.MaskFormerModel

< >

( config: MaskFormerConfig )

Parameters

  • config (MaskFormerConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The bare MaskFormer Model outputting raw hidden-states without any specific head on top. This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< >

( pixel_values: Tensor pixel_mask: typing.Optional[torch.Tensor] = None output_hidden_states: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) → transformers.models.maskformer.modeling_maskformer.MaskFormerModelOutput or tuple(torch.FloatTensor)

Parameters

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Pixel values. Pixel values can be obtained using AutoFeatureExtractor. See AutoFeatureExtractor.__call__() for details.
  • pixel_mask (torch.LongTensor of shape (batch_size, height, width), optional) — Mask to avoid performing attention on padding pixel values. Mask values selected in [0, 1]:

    • 1 for pixels that are real (i.e. not masked),
    • 0 for pixels that are padding (i.e. masked).

    What are attention masks?

  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of Detr’s decoder attention layers.
  • return_dict (bool, optional) — Whether or not to return a ~MaskFormerModelOutput instead of a plain tuple.

Returns

transformers.models.maskformer.modeling_maskformer.MaskFormerModelOutput or tuple(torch.FloatTensor)

A transformers.models.maskformer.modeling_maskformer.MaskFormerModelOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (MaskFormerConfig) and inputs.

  • encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the encoder model (backbone).
  • pixel_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the pixel decoder model (FPN).
  • transformer_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Last hidden states (final feature map) of the last stage of the transformer decoder model.
  • encoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the encoder model at the output of each stage.
  • pixel_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the pixel decoder model at the output of each stage.
  • transformer_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, sequence_length, hidden_size). Hidden-states (also called feature maps) of the transformer decoder at the output of each stage.
  • hidden_states tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor containing encoder_hidden_states, pixel_decoder_hidden_states and decoder_hidden_states
  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length). Attentions weights from Detr’s decoder after the attention softmax, used to compute the weighted average in the self-attention heads.

The MaskFormerModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

>>> from transformers import MaskFormerFeatureExtractor, MaskFormerModel
>>> import torch
>>> from datasets import load_dataset

>>> dataset = load_dataset("huggingface/cats-image")
>>> image = dataset["test"]["image"][0]

>>> feature_extractor = MaskFormerFeatureExtractor.from_pretrained("facebook/maskformer-swin-base-ade")
>>> model = MaskFormerModel.from_pretrained("facebook/maskformer-swin-base-ade")

>>> inputs = feature_extractor(image, return_tensors="pt")

>>> with torch.no_grad():
...     outputs = model(**inputs)

>>> last_hidden_states = outputs.last_hidden_state
>>> list(last_hidden_states.shape)

MaskFormerForInstanceSegmentation

class transformers.MaskFormerForInstanceSegmentation

< >

( config: MaskFormerConfig )

forward

< >

( pixel_values: Tensor mask_labels: typing.Optional[typing.List[torch.Tensor]] = None class_labels: typing.Optional[typing.List[torch.Tensor]] = None pixel_mask: typing.Optional[torch.Tensor] = None output_auxiliary_logits: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) → transformers.models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentationOutput or tuple(torch.FloatTensor)

Parameters

  • pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Pixel values. Pixel values can be obtained using AutoFeatureExtractor. See AutoFeatureExtractor.__call__() for details.
  • pixel_mask (torch.LongTensor of shape (batch_size, height, width), optional) — Mask to avoid performing attention on padding pixel values. Mask values selected in [0, 1]:

    • 1 for pixels that are real (i.e. not masked),
    • 0 for pixels that are padding (i.e. masked).

    What are attention masks?

  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of Detr’s decoder attention layers.
  • return_dict (bool, optional) — Whether or not to return a ~MaskFormerModelOutput instead of a plain tuple.
  • mask_labels (List[torch.Tensor], optional) — List of mask labels of shape (num_labels, height, width) to be fed to a model
  • class_labels (List[torch.LongTensor], optional) — list of target class labels of shape (num_labels, height, width) to be fed to a model. They identify the labels of mask_labels, e.g. the label of mask_labels[i][j] if class_labels[i][j].

Returns

transformers.models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentationOutput or tuple(torch.FloatTensor)

A transformers.models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentationOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (MaskFormerConfig) and inputs.

  • loss (torch.Tensor, optional) — The computed loss, returned when labels are present.
  • class_queries_logits (torch.FloatTensor) — A tensor of shape (batch_size, num_queries, height, width) representing the proposed masks for each query.
  • masks_queries_logits (torch.FloatTensor) — A tensor of shape (batch_size, num_queries, num_labels + 1) representing the proposed classes for each query. Note the + 1 is needed because we incorporate the null class.
  • encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the encoder model (backbone).
  • pixel_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last stage of the pixel decoder model (FPN).
  • transformer_decoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Last hidden states (final feature map) of the last stage of the transformer decoder model.
  • encoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the encoder model at the output of each stage.
  • pixel_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, num_channels, height, width). Hidden-states (also called feature maps) of the pixel decoder model at the output of each stage.
  • transformer_decoder_hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, sequence_length, hidden_size). Hidden-states of the transformer decoder at the output of each stage.
  • hidden_states tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor containing encoder_hidden_states, pixel_decoder_hidden_states and decoder_hidden_states.
  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length). Attentions weights from Detr’s decoder after the attention softmax, used to compute the weighted average in the self-attention heads.

The MaskFormerForInstanceSegmentation forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Examples:

>>> from transformers import MaskFormerFeatureExtractor, MaskFormerForInstanceSegmentation
>>> from PIL import Image
>>> import requests

>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> feature_extractor = MaskFormerFeatureExtractor.from_pretrained("facebook/maskformer-swin-base-ade")
>>> inputs = feature_extractor(images=image, return_tensors="pt")

>>> model = MaskFormerForInstanceSegmentation.from_pretrained("facebook/maskformer-swin-base-ade")
>>> outputs = model(**inputs)
>>> # model predicts class_queries_logits of shape `(batch_size, num_queries)`
>>> # and masks_queries_logits of shape `(batch_size, num_queries, height, width)`
>>> class_queries_logits = outputs.class_queries_logits
>>> masks_queries_logits = outputs.masks_queries_logits

>>> # you can pass them to feature_extractor for postprocessing
>>> output = feature_extractor.post_process_segmentation(outputs)
>>> output = feature_extractor.post_process_semantic_segmentation(outputs)
>>> output = feature_extractor.post_process_panoptic_segmentation(outputs)
←MarianMT M2M100→