easycv.datasets.shared.pipelines package¶

Submodules¶

easycv.datasets.shared.pipelines.dali_transforms module¶

class easycv.datasets.shared.pipelines.dali_transforms.DaliImageDecoder(device='mixed', **kwargs)[source]¶

Bases: object

refer to: https://docs.nvidia.com/deeplearning/dali/archives/dali_0250/user-guide/docs/supported_ops.html#nvidia.dali.ops.ImageDecoder

__init__(device='mixed', **kwargs)[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.dali_transforms.DaliRandomResizedCrop(size, random_area, device='gpu', **kwargs)[source]¶

Bases: object

refer to: https://docs.nvidia.com/deeplearning/dali/archives/dali_0250/user-guide/docs/supported_ops.html#nvidia.dali.ops.RandomResizedCrop

__init__(size, random_area, device='gpu', **kwargs)[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.dali_transforms.DaliResize(resize_shorter, device='gpu', **kwargs)[source]¶

Bases: object

refer to: https://docs.nvidia.com/deeplearning/dali/archives/dali_0250/user-guide/docs/supported_ops.html#nvidia.dali.ops.Resize

__init__(resize_shorter, device='gpu', **kwargs)[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.dali_transforms.DaliColorTwist(prob, saturation, contrast, brightness, hue, device='gpu', center=1)[source]¶

Bases: object

refer to: https://docs.nvidia.com/deeplearning/dali/archives/dali_0250/user-guide/docs/supported_ops.html#nvidia.dali.ops.ColorTwist

__init__(prob, saturation, contrast, brightness, hue, device='gpu', center=1)[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.dali_transforms.DaliRandomGrayscale(prob, device='gpu')[source]¶

Bases: object

refer to: https://docs.nvidia.com/deeplearning/dali/archives/dali_0250/user-guide/docs/supported_ops.html#nvidia.dali.ops.Hsv Create RandomGrayscale op with ops.Hsv. when saturation=0, it represents a grayscale image

__init__(prob, device='gpu')[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.dali_transforms.DaliCropMirrorNormalize(crop, mean, std, prob=0.0, device='gpu', crop_pos_x=0.5, crop_pos_y=0.5, **kwargs)[source]¶

Bases: object

refer to: https://docs.nvidia.com/deeplearning/dali/archives/dali_0250/user-guide/docs/supported_ops.html#nvidia.dali.ops.CropMirrorNormalize

__init__(crop, mean, std, prob=0.0, device='gpu', crop_pos_x=0.5, crop_pos_y=0.5, **kwargs)[source]¶: Initialize self. See help(type(self)) for accurate signature.

easycv.datasets.shared.pipelines.format module¶

easycv.datasets.shared.pipelines.format.to_tensor(data)[source]¶

Convert objects of various python types to torch.Tensor.

Supported types are: numpy.ndarray, torch.Tensor, Sequence, int and float.

Parameters: data (torch.Tensor | numpy.ndarray | Sequence | int | float) – Data to be converted.

class easycv.datasets.shared.pipelines.format.ImageToTensor(keys)[source]¶

Bases: object

Convert image to torch.Tensor by given keys.

The dimension order of input image is (H, W, C). The pipeline will convert it to (C, H, W). If only 2 dimension (H, W) is given, the output would be (1, H, W).

Parameters: keys (Sequence[str]) – Key of images to be converted to Tensor.

__init__(keys)[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.format.Collect(keys, meta_keys=('filename', 'ori_filename', 'ori_img_shape', 'img_shape', 'scale_factor', 'pad', 'flip', 'flip_direction', 'img_norm_cfg'))[source]¶

Bases: object

Collect data from the loader relevant to the specific task.

This is usually the last stage of the data loader pipeline. Typically keys is set to some subset of “img”, “proposals”, “gt_bboxes”, “gt_bboxes_ignore”, “gt_labels”, and/or “gt_masks”.

The “img_meta” item is always populated. The contents of the “img_meta” dictionary depends on “meta_keys”. By default this includes:

“img_shape”: shape of the image input to the network as a tuple (h, w). Note that images may be zero padded on the bottom/right if the batch tensor is larger than this shape.

“scale_factor”: a float indicating the preprocessing scale

“flip”: a boolean indicating if image flip transform was used

“filename”: path to the image file

“ori_img_shape”: original shape of the image as a tuple (h, w, c)

“img_norm_cfg”: a dict of normalization information:

mean - per channel mean subtraction

std - per channel std divisor

to_rgb - bool indicating if bgr was converted to rgb

Parameters

keys (Sequence[str]) – Keys of results to be collected in data.
meta_keys (Sequence[str], optional) –
Meta keys to be converted to mmcv.DataContainer and collected in data[img_metas]. Default: ``(‘filename’, ‘ori_filename’, ‘ori_img_shape’, ‘img_shape’,

’scale_factor’, ‘flip’, ‘flip_direction’, ‘img_norm_cfg’)``

__init__(keys, meta_keys=('filename', 'ori_filename', 'ori_img_shape', 'img_shape', 'scale_factor', 'pad', 'flip', 'flip_direction', 'img_norm_cfg'))[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.format.DefaultFormatBundle[source]¶

Bases: object

Default formatting bundle. It simplifies the pipeline of formatting common fields, including “img”, “proposals”, “gt_bboxes”, “gt_labels”, “gt_masks” and “gt_semantic_seg”. These fields are formatted as follows. - img: (1)transpose, (2)to tensor, (3)to DataContainer (stack=True) - proposals: (1)to tensor, (2)to DataContainer - gt_bboxes: (1)to tensor, (2)to DataContainer - gt_bboxes_ignore: (1)to tensor, (2)to DataContainer - gt_labels: (1)to tensor, (2)to DataContainer - gt_masks: (1)to tensor, (2)to DataContainer (cpu_only=True) - gt_semantic_seg: (1)unsqueeze dim-0 (2)to tensor, (3)to DataContainer (stack=True)

easycv.datasets.shared.pipelines.third_transforms_wrapper module¶

easycv.datasets.shared.pipelines.third_transforms_wrapper.is_child_of(obj, cls)[source]¶

easycv.datasets.shared.pipelines.third_transforms_wrapper.get_args(obj)[source]¶

easycv.datasets.shared.pipelines.third_transforms_wrapper.wrap_torchvision_transforms(transform_obj)[source]¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.AugMix(severity: int = 3, mixture_width: int = 3, chain_depth: int = -1, alpha: float = 1.0, all_ops: bool = True, interpolation: torchvision.transforms.functional.InterpolationMode = <InterpolationMode.BILINEAR: 'bilinear'>, fill: Optional[List[float]] = None)¶

Bases: torchvision.transforms.autoaugment.AugMix

forward(results)¶

img (PIL Image or Tensor): Image to be transformed.

Returns: Transformed image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.AutoAugment(policy: torchvision.transforms.autoaugment.AutoAugmentPolicy = <AutoAugmentPolicy.IMAGENET: 'imagenet'>, interpolation: torchvision.transforms.functional.InterpolationMode = <InterpolationMode.NEAREST: 'nearest'>, fill: Optional[List[float]] = None)¶

Bases: torchvision.transforms.autoaugment.AutoAugment

forward(results)¶

img (PIL Image or Tensor): Image to be transformed.

Returns: AutoAugmented image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.CenterCrop(size)¶

Bases: torchvision.transforms.transforms.CenterCrop

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be cropped.
Returns: Cropped image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.ColorJitter(brightness=0, contrast=0, saturation=0, hue=0)¶

Bases: torchvision.transforms.transforms.ColorJitter

forward(results)¶

Parameters: img (PIL Image or Tensor) – Input image.
Returns: Color jittered image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.ConvertImageDtype(dtype: torch.dtype)¶

Bases: torchvision.transforms.transforms.ConvertImageDtype

forward(results)¶

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

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 registered hooks while the latter silently ignores them.

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.ElasticTransform(alpha=50.0, sigma=5.0, interpolation=<InterpolationMode.BILINEAR: 'bilinear'>, fill=0)¶

Bases: torchvision.transforms.transforms.ElasticTransform

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be transformed.
Returns: Transformed image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.FiveCrop(size)¶

Bases: torchvision.transforms.transforms.FiveCrop

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be cropped.
Returns: tuple of 5 images. Image can be PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.GaussianBlur(kernel_size, sigma=(0.1, 2.0))¶

Bases: torchvision.transforms.transforms.GaussianBlur

forward(results)¶

Parameters: img (PIL Image or Tensor) – image to be blurred.
Returns: Gaussian blurred image
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.Grayscale(num_output_channels=1)¶

Bases: torchvision.transforms.transforms.Grayscale

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be converted to grayscale.
Returns: Grayscaled image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.Lambda(lambd)¶: Bases: torchvision.transforms.transforms.Lambda

class easycv.datasets.shared.pipelines.third_transforms_wrapper.LinearTransformation(transformation_matrix, mean_vector)¶

Bases: torchvision.transforms.transforms.LinearTransformation

forward(results)¶

Parameters: tensor (Tensor) – Tensor image to be whitened.
Returns: Transformed image.
Return type: Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.Normalize(mean, std, inplace=False)¶

Bases: torchvision.transforms.transforms.Normalize

forward(results)¶

Parameters: tensor (Tensor) – Tensor image to be normalized.
Returns: Normalized Tensor image.
Return type: Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.PILToTensor¶: Bases: torchvision.transforms.transforms.PILToTensor

class easycv.datasets.shared.pipelines.third_transforms_wrapper.Pad(padding, fill=0, padding_mode='constant')¶

Bases: torchvision.transforms.transforms.Pad

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be padded.
Returns: Padded image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandAugment(num_ops: int = 2, magnitude: int = 9, num_magnitude_bins: int = 31, interpolation: torchvision.transforms.functional.InterpolationMode = <InterpolationMode.NEAREST: 'nearest'>, fill: Optional[List[float]] = None)¶

Bases: torchvision.transforms.autoaugment.RandAugment

forward(results)¶

img (PIL Image or Tensor): Image to be transformed.

Returns: Transformed image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomAdjustSharpness(sharpness_factor, p=0.5)¶

Bases: torchvision.transforms.transforms.RandomAdjustSharpness

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be sharpened.
Returns: Randomly sharpened image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomAffine(degrees, translate=None, scale=None, shear=None, interpolation=<InterpolationMode.NEAREST: 'nearest'>, fill=0, center=None)¶

Bases: torchvision.transforms.transforms.RandomAffine

forward(results)¶

img (PIL Image or Tensor): Image to be transformed.

Returns: Affine transformed image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomAutocontrast(p=0.5)¶

Bases: torchvision.transforms.transforms.RandomAutocontrast

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be autocontrasted.
Returns: Randomly autocontrasted image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomChoice(transforms, p=None)¶: Bases: torchvision.transforms.transforms.RandomChoice

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomCrop(size, padding=None, pad_if_needed=False, fill=0, padding_mode='constant')¶

Bases: torchvision.transforms.transforms.RandomCrop

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be cropped.
Returns: Cropped image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomEqualize(p=0.5)¶

Bases: torchvision.transforms.transforms.RandomEqualize

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be equalized.
Returns: Randomly equalized image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomErasing(p=0.5, scale=(0.02, 0.33), ratio=(0.3, 3.3), value=0, inplace=False)¶

Bases: torchvision.transforms.transforms.RandomErasing

forward(results)¶

Parameters: img (Tensor) – Tensor image to be erased.
Returns: Erased Tensor image.
Return type: img (Tensor)

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomGrayscale(p=0.1)¶

Bases: torchvision.transforms.transforms.RandomGrayscale

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be converted to grayscale.
Returns: Randomly grayscaled image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomHorizontalFlip(p=0.5)¶

Bases: torchvision.transforms.transforms.RandomHorizontalFlip

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be flipped.
Returns: Randomly flipped image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomInvert(p=0.5)¶

Bases: torchvision.transforms.transforms.RandomInvert

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be inverted.
Returns: Randomly color inverted image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomOrder(transforms)¶: Bases: torchvision.transforms.transforms.RandomOrder

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomPerspective(distortion_scale=0.5, p=0.5, interpolation=<InterpolationMode.BILINEAR: 'bilinear'>, fill=0)¶

Bases: torchvision.transforms.transforms.RandomPerspective

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be Perspectively transformed.
Returns: Randomly transformed image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomPosterize(bits, p=0.5)¶

Bases: torchvision.transforms.transforms.RandomPosterize

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be posterized.
Returns: Randomly posterized image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomResizedCrop(size, scale=(0.08, 1.0), ratio=(0.75, 1.3333333333333333), interpolation=<InterpolationMode.BILINEAR: 'bilinear'>, antialias: Optional[bool] = None)¶

Bases: torchvision.transforms.transforms.RandomResizedCrop

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be cropped and resized.
Returns: Randomly cropped and resized image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomRotation(degrees, interpolation=<InterpolationMode.NEAREST: 'nearest'>, expand=False, center=None, fill=0)¶

Bases: torchvision.transforms.transforms.RandomRotation

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be rotated.
Returns: Rotated image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomSolarize(threshold, p=0.5)¶

Bases: torchvision.transforms.transforms.RandomSolarize

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be solarized.
Returns: Randomly solarized image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.RandomVerticalFlip(p=0.5)¶

Bases: torchvision.transforms.transforms.RandomVerticalFlip

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be flipped.
Returns: Randomly flipped image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.Resize(size, interpolation=<InterpolationMode.BILINEAR: 'bilinear'>, max_size=None, antialias=None)¶

Bases: torchvision.transforms.transforms.Resize

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be scaled.
Returns: Rescaled image.
Return type: PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.TenCrop(size, vertical_flip=False)¶

Bases: torchvision.transforms.transforms.TenCrop

forward(results)¶

Parameters: img (PIL Image or Tensor) – Image to be cropped.
Returns: tuple of 10 images. Image can be PIL Image or Tensor

training: bool¶

class easycv.datasets.shared.pipelines.third_transforms_wrapper.ToPILImage(mode=None)¶: Bases: torchvision.transforms.transforms.ToPILImage

class easycv.datasets.shared.pipelines.third_transforms_wrapper.ToTensor¶: Bases: torchvision.transforms.transforms.ToTensor

class easycv.datasets.shared.pipelines.third_transforms_wrapper.TrivialAugmentWide(num_magnitude_bins: int = 31, interpolation: torchvision.transforms.functional.InterpolationMode = <InterpolationMode.NEAREST: 'nearest'>, fill: Optional[List[float]] = None)¶

Bases: torchvision.transforms.autoaugment.TrivialAugmentWide

forward(results)¶

img (PIL Image or Tensor): Image to be transformed.

Returns: Transformed image.
Return type: PIL Image or Tensor

training: bool¶

easycv.datasets.shared.pipelines.transforms module¶

class easycv.datasets.shared.pipelines.transforms.Compose(transforms, profiling=False)[source]¶

Bases: object

Compose a data pipeline with a sequence of transforms. :param transforms: Either config dicts of transforms or transform objects. :type transforms: list[dict | callable]

__init__(transforms, profiling=False)[source]¶: Initialize self. See help(type(self)) for accurate signature.

class easycv.datasets.shared.pipelines.transforms.LoadImage(to_float32=False, mode='bgr')[source]¶

Bases: object

Load an image from file or numpy or PIL object. :param to_float32: Whether to convert the loaded image to a float32

numpy array. If set to False, the loaded image is an uint8 array. Defaults to False.

__init__(to_float32=False, mode='bgr')[source]¶: Initialize self. See help(type(self)) for accurate signature.