vision API reference
186 public APIs from vision (pytorch/vision) — 109 classes, 62 functions, 15 methods. Signatures extracted by static analysis of the actual source.
Repository: pytorch/vision
| Kind | Count |
|---|---|
| Classes | 109 |
| Functions | 62 |
| Methods | 15 |
API list
class
references.classification.transforms.RandomCutMixRandomly apply CutMix to the provided batch and targets.
class
references.classification.transforms.RandomMixUpRandomly apply MixUp to the provided batch and targets.
func
references.depth.stereo.train.shuffle_dataset(dataset)Shuffle the dataset
func
references.depth.stereo.utils.losses.make_gaussian_kernel(kernel_size:int, sigma:float) -> torch.TensorFunction to create a 2D Gaussian kernel.
class
references.depth.stereo.utils.padder.InputPadderPads images such that dimensions are divisible by 8
class
references.optical_flow.utils.InputPadderPads images such that dimensions are divisible by 8
class
references.video_classification.transforms.ConvertBCHWtoCBHWConvert tensor from (B, C, H, W) to (C, B, H, W)
class
torchvision.datasets._stereo_matching.StereoMatchingDatasetBase interface for Stereo matching datasets
class
torchvision.datasets.folder.DatasetFolderA generic data loader.
func
torchvision.datasets.folder.find_classes(directory:Union[str, Path]) -> tuple[list[str], dict[str, int]]Finds the class folders in a dataset.
func
torchvision.datasets.folder.has_file_allowed_extension(filename:str, extensions:Union[str, tuple[str, ...]]) -> boolChecks if a file is an allowed extension.
func
torchvision.datasets.folder.is_image_file(filename:str) -> boolChecks if a file is an allowed image extension.
class
torchvision.datasets.lfw.LFWPairs`LFW <http://vis-www.cs.umass.edu/lfw/>`_ Dataset.
class
torchvision.datasets.lfw.LFWPeople`LFW <http://vis-www.cs.umass.edu/lfw/>`_ Dataset.
class
torchvision.datasets.lsun.LSUN`LSUN <https://paperswithcode.com/dataset/lsun>`_ dataset.
class
torchvision.datasets.mnist.MNIST`MNIST <http://yann.lecun.com/exdb/mnist/>`_ Dataset.
method
torchvision.datasets.mnist.MNIST.download() -> NoneDownload the MNIST data if it doesn't exist already.
method
torchvision.datasets.mnist.QMNIST.download() -> NoneDownload the QMNIST data if it doesn't exist already.
class
torchvision.datasets.pcam.PCAM`PCAM Dataset <https://github.com/basveeling/pcam>`_.
func
torchvision.datasets.phototour.PIL2array(_img:Image.Image) -> np.ndarrayConvert PIL image type to numpy 2D array
func
torchvision.datasets.phototour.read_image_file(data_dir:str, image_ext:str, n:int) -> torch.TensorReturn a Tensor containing the patches
class
torchvision.datasets.stl10.STL10`STL10 <https://cs.stanford.edu/~acoates/stl10/>`_ Dataset.
class
torchvision.datasets.svhn.SVHN`SVHN <http://ufldl.stanford.edu/housenumbers/>`_ Dataset.
class
torchvision.datasets.ucf101.UCF101`UCF101 <https://www.crcv.ucf.edu/data/UCF101.php>`_ dataset.
func
torchvision.datasets.utils.extract_archive(from_path:Union[str, pathlib.Path], to_path:Optional[Union[str, pathlib.Path]]=None, remove_finished:bool=False) -> Union[str, pathlib.Path]Extract an archive.
class
torchvision.datasets.vision.VisionDatasetBase Class For making datasets which are compatible with torchvision.
class
torchvision.datasets.widerface.WIDERFace`WIDERFace <http://shuoyang1213.me/WIDERFACE/>`_ Dataset.
class
torchvision.io.image.ImageReadModeAllow automatic conversion to RGB, RGBA, etc while decoding.
func
torchvision.io.image.decode_avif(input:torch.Tensor, mode:ImageReadMode=ImageReadMode.UNCHANGED) -> torch.TensorDecode an AVIF image into a 3 dimensional RGB[A] Tensor.
func
torchvision.io.image.decode_gif(input:torch.Tensor) -> torch.TensorDecode a GIF image into a 3 or 4 dimensional RGB Tensor.
func
torchvision.io.image.decode_heic(input:torch.Tensor, mode:ImageReadMode=ImageReadMode.UNCHANGED) -> torch.TensorDecode an HEIC image into a 3 dimensional RGB[A] Tensor.
func
torchvision.io.image.decode_webp(input:torch.Tensor, mode:ImageReadMode=ImageReadMode.UNCHANGED) -> torch.TensorDecode a WEBP image into a 3 dimensional RGB[A] Tensor.
func
torchvision.io.image.read_file(path:str) -> torch.TensorReturn the bytes contents of a file as a uint8 1D Tensor.
func
torchvision.io.image.write_file(filename:str, data:torch.Tensor) -> torch.TensorWrite the content of an uint8 1D tensor to a file.
class
torchvision.models._api.WeightsEnumThis class is the parent class of all model weights.
func
torchvision.models._api.get_model_builder(name:str) -> Callable[..., nn.Module]Gets the model name and returns the model builder method.
func
torchvision.models._api.get_model_weights(name:Union[Callable, str]) -> type[WeightsEnum]Returns the weights enum class associated to the given model.
func
torchvision.models._api.get_weight(name:str) -> WeightsEnumGets the weights enum value by its full name.
class
torchvision.models.detection._utils.BoxLinearCoderThe linear box-to-box transform defined in FCOS.
class
torchvision.models.detection.backbone_utils.BackboneWithFPNAdds a FPN on top of a model.
class
torchvision.models.detection.faster_rcnn.FasterRCNNImplements Faster R-CNN.
class
torchvision.models.detection.fcos.FCOSImplements FCOS.
class
torchvision.models.detection.fcos.FCOSClassificationHeadA classification head for use in FCOS.
class
torchvision.models.detection.fcos.FCOSHeadA regression and classification head for use in FCOS.
class
torchvision.models.detection.generalized_rcnn.GeneralizedRCNNMain class for Generalized R-CNN.
class
torchvision.models.detection.keypoint_rcnn.KeypointRCNNImplements Keypoint R-CNN.
class
torchvision.models.detection.mask_rcnn.MaskRCNNImplements Mask R-CNN.
class
torchvision.models.detection.retinanet.RetinaNetImplements RetinaNet.
class
torchvision.models.detection.retinanet.RetinaNetClassificationHeadA classification head for use in RetinaNet.
class
torchvision.models.detection.retinanet.RetinaNetRegressionHeadA regression head for use in RetinaNet.
class
torchvision.models.detection.rpn.RegionProposalNetworkImplements Region Proposal Network (RPN).
class
torchvision.models.feature_extraction.DualGraphModuleA derivative of `fx.GraphModule`.
class
torchvision.models.maxvit.MBConvMBConv: Mobile Inverted Residual Bottleneck.
method
torchvision.models.maxvit.MBConv.forward(x:Tensor) -> TensorArgs: x (Tensor): Input tensor with expected layout of [B, C, H, W].
class
torchvision.models.maxvit.MaxVitBlockA MaxVit block consisting of `n_layers` MaxVit layers.
method
torchvision.models.maxvit.MaxVitBlock.forward(x:Tensor) -> TensorArgs: x (Tensor): Input tensor of shape (B, C, H, W).
class
torchvision.models.maxvit.RelativePositionalMultiHeadAttentionRelative Positional Multi-Head Attention.
class
torchvision.models.maxvit.SwapAxesPermute the axes of a tensor.
class
torchvision.models.maxvit.WindowPartitionPartition the input tensor into non-overlapping windows.
class
torchvision.models.mnasnet.MNASNetMNASNet, as described in https://arxiv.org/abs/1807.11626.
class
torchvision.models.optical_flow.raft.ConvGRUConvolutional Gru unit.
class
torchvision.models.optical_flow.raft.CorrBlockThe correlation block.
class
torchvision.models.optical_flow.raft.FlowHeadFlow head, part of the update block.
class
torchvision.models.optical_flow.raft.MotionEncoderThe motion encoder, part of the update block.
class
torchvision.models.optical_flow.raft.Raft_Large_WeightsThe metrics reported here are as follows.
class
torchvision.models.optical_flow.raft.Raft_Small_WeightsThe metrics reported here are as follows.
class
torchvision.models.optical_flow.raft.RecurrentBlockRecurrent block, part of the update block.
class
torchvision.models.regnet.AnyStageAnyNet stage (sequence of blocks w/ the same output shape).
class
torchvision.models.regnet.BottleneckTransformBottleneck transformation: 1x1, 3x3 [+SE], 1x1.
class
torchvision.models.regnet.SimpleStemINSimple stem for ImageNet: 3x3, BN, ReLU.
func
torchvision.models.resnet.conv1x1(in_planes:int, out_planes:int, stride:int=1) -> nn.Conv2d1x1 convolution
func
torchvision.models.resnet.conv3x3(in_planes:int, out_planes:int, stride:int=1, groups:int=1, dilation:int=1) -> nn.Conv2d3x3 convolution with padding
class
torchvision.models.swin_transformer.PatchMergingPatch Merging Layer.
class
torchvision.models.swin_transformer.PatchMergingV2Patch Merging Layer for Swin Transformer V2.
class
torchvision.models.swin_transformer.ShiftedWindowAttentionSee :func:`shifted_window_attention`.
class
torchvision.models.swin_transformer.ShiftedWindowAttentionV2See :func:`shifted_window_attention_v2`.
class
torchvision.models.swin_transformer.SwinTransformerBlockSwin Transformer Block.
class
torchvision.models.swin_transformer.SwinTransformerBlockV2Swin Transformer V2 Block.
class
torchvision.models.video.resnet.BasicStemThe default conv-batchnorm-relu stem
func
torchvision.models.video.resnet.mc3_18(*weights:Optional[MC3_18_Weights]=None, *progress:bool=True, **kwargs:Any) -> VideoResNetConstruct 18 layer Mixed Convolution network as in ..
func
torchvision.models.video.resnet.r2plus1d_18(*weights:Optional[R2Plus1D_18_Weights]=None, *progress:bool=True, **kwargs:Any) -> VideoResNetConstruct 18 layer deep R(2+1)D network as in ..
func
torchvision.models.video.resnet.r3d_18(*weights:Optional[R3D_18_Weights]=None, *progress:bool=True, **kwargs:Any) -> VideoResNetConstruct 18 layer Resnet3D model.
class
torchvision.models.video.s3d.S3DS3D main class.
func
torchvision.models.video.s3d.s3d(*weights:Optional[S3D_Weights]=None, *progress:bool=True, **kwargs:Any) -> S3DConstruct Separable 3D CNN model.
class
torchvision.models.video.swin_transformer.PatchEmbed3dVideo to Patch Embedding.
method
torchvision.models.video.swin_transformer.PatchEmbed3d.forward(x:Tensor) -> TensorForward function.
class
torchvision.models.video.swin_transformer.ShiftedWindowAttention3dSee :func:`shifted_window_attention_3d`.
class
torchvision.models.vision_transformer.EncoderBlockTransformer encoder block.
class
torchvision.models.vision_transformer.MLPBlockTransformer MLP block.
func
torchvision.ops.boxes.batched_nms(boxes:Tensor, scores:Tensor, idxs:Tensor, iou_threshold:float) -> TensorPerforms non-maximum suppression in a batched fashion.
func
torchvision.ops.boxes.box_area(boxes:Tensor, fmt:str='xyxy') -> TensorComputes the area of a set of bounding boxes from a given format.
func
torchvision.ops.boxes.clip_boxes_to_image(boxes:Tensor, size:tuple[int, int]) -> TensorClip boxes so that they lie inside an image of size ``size``.
func
torchvision.ops.boxes.masks_to_boxes(masks:torch.Tensor) -> torch.TensorCompute the bounding boxes around the provided masks.
class
torchvision.ops.deform_conv.DeformConv2dSee :func:`deform_conv2d`.
class
torchvision.ops.drop_block.DropBlock2dSee :func:`drop_block2d`.
class
torchvision.ops.drop_block.DropBlock3dSee :func:`drop_block3d`.
class
torchvision.ops.feature_pyramid_network.ExtraFPNBlockBase class for the extra block in the FPN.
class
torchvision.ops.misc.MLPThis block implements the multi-layer perceptron (MLP) module.
class
torchvision.ops.ps_roi_align.PSRoIAlignSee :func:`ps_roi_align`.
class
torchvision.ops.ps_roi_pool.PSRoIPoolSee :func:`ps_roi_pool`.
class
torchvision.ops.roi_align.RoIAlignSee :func:`roi_align`.
class
torchvision.ops.roi_pool.RoIPoolSee :func:`roi_pool`.
class
torchvision.ops.stochastic_depth.StochasticDepthSee :func:`stochastic_depth`.
method
torchvision.transforms.autoaugment.AugMix.forward(orig_img:Tensor) -> Tensorimg (PIL Image or Tensor): Image to be transformed.
class
torchvision.transforms.autoaugment.AutoAugmentPolicyAutoAugment policies learned on different datasets.
func
torchvision.transforms.functional.adjust_brightness(img:Tensor, brightness_factor:float) -> TensorAdjust brightness of an image.
func
torchvision.transforms.functional.adjust_contrast(img:Tensor, contrast_factor:float) -> TensorAdjust contrast of an image.
func
torchvision.transforms.functional.adjust_gamma(img:Tensor, gamma:float, gain:float=1) -> TensorPerform gamma correction on an image.
func
torchvision.transforms.functional.adjust_hue(img:Tensor, hue_factor:float) -> TensorAdjust hue of an image.
func
torchvision.transforms.functional.adjust_saturation(img:Tensor, saturation_factor:float) -> TensorAdjust color saturation of an image.
func
torchvision.transforms.functional.adjust_sharpness(img:Tensor, sharpness_factor:float) -> TensorAdjust the sharpness of an image.
func
torchvision.transforms.functional.center_crop(img:Tensor, output_size:list[int]) -> TensorCrops the given image at the center.
func
torchvision.transforms.functional.crop(img:Tensor, top:int, left:int, height:int, width:int) -> TensorCrop the given image at specified location and output size.
func
torchvision.transforms.functional.erase(img:Tensor, i:int, j:int, h:int, w:int, v:Tensor, inplace:bool=False) -> TensorErase the input Tensor Image with given value.
func
torchvision.transforms.functional.get_dimensions(img:Tensor) -> list[int]Returns the dimensions of an image as [channels, height, width].
func
torchvision.transforms.functional.get_image_num_channels(img:Tensor) -> intReturns the number of channels of an image.
func
torchvision.transforms.functional.get_image_size(img:Tensor) -> list[int]Returns the size of an image as [width, height].
func
torchvision.transforms.functional.hflip(img:Tensor) -> TensorHorizontally flip the given image.
func
torchvision.transforms.functional.invert(img:Tensor) -> TensorInvert the colors of an RGB/grayscale image.
func
torchvision.transforms.functional.pil_to_tensor(pic:Any) -> TensorConvert a ``PIL Image`` to a tensor of the same type.
func
torchvision.transforms.functional.rgb_to_grayscale(img:Tensor, num_output_channels:int=1) -> TensorConvert RGB image to grayscale version of image.
func
torchvision.transforms.functional.rotate(img:Tensor, angle:float, interpolation:InterpolationMode=InterpolationMode.NEAREST, expand:bool=False, center:Optional[list[int]]=None, fill:Optional[list[float]]=None) -> TensorRotate the image by angle.
func
torchvision.transforms.functional.to_tensor(pic:Union[PILImage, np.ndarray]) -> TensorConvert a ``PIL Image`` or ``numpy.ndarray`` to tensor.
func
torchvision.transforms.functional.vflip(img:Tensor) -> TensorVertically flip the given image.
class
torchvision.transforms.transforms.CenterCropCrops the given image at the center.
class
torchvision.transforms.transforms.ComposeComposes several transforms together.
class
torchvision.transforms.transforms.ElasticTransformTransform a tensor image with elastic transformations.
method
torchvision.transforms.transforms.ElasticTransform.forward(tensor:Tensor) -> TensorArgs: tensor (PIL Image or Tensor): Image to be transformed.
class
torchvision.transforms.transforms.GaussianBlurBlurs image with randomly chosen Gaussian blur.
method
torchvision.transforms.transforms.GaussianBlur.forward(img:Tensor) -> TensorArgs: img (PIL Image or Tensor): image to be blurred.
method
torchvision.transforms.transforms.GaussianBlur.get_params(sigma_min:float, sigma_max:float) -> floatChoose sigma for random gaussian blurring.
class
torchvision.transforms.transforms.GrayscaleConvert image to grayscale.
class
torchvision.transforms.transforms.LambdaApply a user-defined lambda as a transform.
class
torchvision.transforms.transforms.NormalizeNormalize a tensor image with mean and standard deviation.
method
torchvision.transforms.transforms.Normalize.forward(tensor:Tensor) -> TensorArgs: tensor (Tensor): Tensor image to be normalized.
class
torchvision.transforms.transforms.PadPad the given image on all sides with the given "pad" value.
class
torchvision.transforms.transforms.RandomChoiceApply single transformation randomly picked from a list.
class
torchvision.transforms.transforms.RandomCropCrop the given image at a random location.
method
torchvision.transforms.transforms.RandomCrop.get_params(img:Tensor, output_size:tuple[int, int]) -> tuple[int, int, int, int]Get parameters for ``crop`` for a random crop.
class
torchvision.transforms.transforms.RandomOrderApply a list of transformations in a random order.
class
torchvision.transforms.transforms.RandomRotationRotate the image by angle.
method
torchvision.transforms.transforms.RandomRotation.get_params(degrees:list[float]) -> floatGet parameters for ``rotate`` for a random rotation.
class
torchvision.transforms.transforms.ResizeResize the input image to the given size.
class
torchvision.transforms.v2._augment.CutMixApply CutMix to the provided batch of images and labels.
class
torchvision.transforms.v2._augment.MixUpApply MixUp to the provided batch of images and labels.
class
torchvision.transforms.v2._color.GrayscaleConvert images or videos to grayscale.
class
torchvision.transforms.v2._color.RandomChannelPermutationRandomly permute the channels of an image or video
class
torchvision.transforms.v2._container.ComposeComposes several transforms together.
class
torchvision.transforms.v2._container.RandomChoiceApply single transformation randomly picked from a list.
class
torchvision.transforms.v2._container.RandomOrderApply a list of transformations in a random order.
class
torchvision.transforms.v2._geometry.CenterCropCrop the input at the center.
class
torchvision.transforms.v2._geometry.ElasticTransformTransform the input with elastic transformations.
class
torchvision.transforms.v2._geometry.PadPad the input on all sides with the given "pad" value.
class
torchvision.transforms.v2._geometry.RandomCropCrop the input at a random location.
class
torchvision.transforms.v2._geometry.RandomHorizontalFlipHorizontally flip the input with a given probability.
class
torchvision.transforms.v2._geometry.RandomResizeRandomly resize the input.
class
torchvision.transforms.v2._geometry.RandomRotationRotate the input by angle.
class
torchvision.transforms.v2._geometry.RandomShortestSizeRandomly resize the input.
class
torchvision.transforms.v2._geometry.RandomVerticalFlipVertically flip the input with a given probability.
class
torchvision.transforms.v2._geometry.ResizeResize the input to the given size.
class
torchvision.transforms.v2._meta.ClampKeyPointsClamp keypoints to their corresponding image dimensions.
class
torchvision.transforms.v2._misc.GaussianBlurBlurs image with randomly chosen Gaussian blur kernel.
class
torchvision.transforms.v2._misc.GaussianNoiseAdd gaussian noise to images or videos.
class
torchvision.transforms.v2._misc.LambdaApply a user-defined function as a transform.
class
torchvision.transforms.v2._transform.TransformBase class to implement your own v2 transforms.
method
torchvision.transforms.v2._transform.Transform.forward(*inputs:Any) -> AnyDo not override this!
method
torchvision.transforms.v2._transform.Transform.make_params(flat_inputs:list[Any]) -> dict[str, Any]Method to override for custom transforms.
method
torchvision.transforms.v2._transform.Transform.transform(inpt:Any, params:dict[str, Any]) -> AnyMethod to override for custom transforms.
func
torchvision.transforms.v2._utils.get_bounding_boxes(flat_inputs:list[Any]) -> tv_tensors.BoundingBoxesReturn the Bounding Boxes in the input.
func
torchvision.transforms.v2._utils.get_keypoints(flat_inputs:list[Any]) -> tv_tensors.KeyPointsReturn the keypoints in the input.
func
torchvision.transforms.v2._utils.query_chw(flat_inputs:list[Any]) -> tuple[int, int, int]Return Channel, Height, and Width.
func
torchvision.transforms.v2._utils.query_size(flat_inputs:list[Any]) -> tuple[int, int]Return Height and Width.
func
torchvision.transforms.v2.functional._augment.jpeg(image:torch.Tensor, quality:int) -> torch.TensorSee :class:`~torchvision.transforms.v2.JPEG` for details.
func
torchvision.transforms.v2.functional._color.adjust_brightness(inpt:torch.Tensor, brightness_factor:float) -> torch.TensorAdjust brightness.
func
torchvision.transforms.v2.functional._color.adjust_gamma(inpt:torch.Tensor, gamma:float, gain:float=1) -> torch.TensorAdjust gamma.
func
torchvision.transforms.v2.functional._color.adjust_hue(inpt:torch.Tensor, hue_factor:float) -> torch.TensorAdjust hue
func
torchvision.transforms.v2.functional._color.adjust_saturation(inpt:torch.Tensor, saturation_factor:float) -> torch.TensorAdjust saturation.
func
torchvision.transforms.v2.functional._color.grayscale_to_rgb(inpt:torch.Tensor) -> torch.TensorSee :class:`~torchvision.transforms.v2.RGB` for details.
func
torchvision.transforms.v2.functional._color.invert(inpt:torch.Tensor) -> torch.TensorSee :func:`~torchvision.transforms.v2.RandomInvert`.
func
torchvision.transforms.v2.functional._deprecated.to_tensor(inpt:Any) -> torch.Tensor[DEPREACTED] Use to_image() and to_dtype() instead.
func
torchvision.transforms.v2.functional._misc.convert_image_dtype(image:torch.Tensor, dtype:torch.dtype=torch.float32) -> torch.Tensor[DEPRECATED] Use to_dtype() instead.
class
torchvision.tv_tensors._bounding_boxes.BoundingBoxFormatCoordinate format of a bounding box.
class
torchvision.tv_tensors._tv_tensor.TVTensorBase class for all TVTensors.
func
torchvision.utils.flow_to_image(flow:torch.Tensor) -> torch.TensorConverts a flow to an RGB image.
func
torchvision.utils.make_grid(tensor:Union[torch.Tensor, list[torch.Tensor]], nrow:int=8, padding:int=2, normalize:bool=False, value_range:Optional[tuple[int, int]]=None, scale_each:bool=False, pad_value:float=0.0) -> torch.TensorMake a grid of images.
func
torchvision.utils.save_image(tensor:Union[torch.Tensor, list[torch.Tensor]], fp:Union[str, pathlib.Path, BinaryIO], format:Optional[str]=None, **kwargs) -> NoneSave a given Tensor into an image file.
About this data
These signatures were extracted from the public source of pytorch/vision
using Python's ast module. Argument names, default values,
type annotations and return types are taken verbatim from the code.
Implementation bodies are never stored. See
how it works for details.