mindvideo.data.transforms

VideoCenterCrop

class mindvideo.data.transforms.VideoCenterCrop(size=(224, 224)):

Crop each frame of the input video at the center to the given size. If input frame of video size is smaller than output size, input video will be padded with 0 before cropping.

  • base: trans.PyTensorOperation

Parameters:

  • size (Union[int, sequence]): The output size of the cropped image. If size is an integer, a square crop of size (size, size) is returned. If size is a sequence of length 2, it should be (height, width).Default:(224,224)

Return:

None

VideoNormalize

class mindvideo.data.transforms.VideoNormalize(mean, std):

VideoNormalize the input numpy.ndarray video of shape (C, T, H, W) with the specified mean and standard deviation.

  • base: trans.PyTensorOperation

Note:

The values of the input image need to be in the range [0.0, 1.0]. If not so, call VideoReOrder and VideoRescale first.

Parameters:

  • mean (Union[float, sequence]): list or tuple of mean values for each channel, arranged in channel order. The values must be in the range [0.0, 1.0]. If a single float is provided, it will be filled to the same length as the channel.

  • std (Union[float, sequence]): list or tuple of standard deviation values for each channel, arranged in channel order. The values must be in the range (0.0, 1.0]. If a single float is provided, it will be filled to the same length as the channel.

Return:

None

VideoRandomCrop

class mindvideo.data.transforms.VideoRandomCrop(size):

Crop the given video sequences (t x h x w x c) at a random location.

Parameters:

  • size (sequence or int): Desired output size of the crop. If size is an int instead of sequence like (h, w), a square crop (size, size) is made.

Return:

None

def mindvideo.data.transforms.VideoNormalize.get_params(img, output_size)

Get parameters for crop for a random crop.

Parameters:

  • img (PIL Image): Image to be cropped.

  • output_size (tuple): Expected output size of the crop.

Return:

  • tuple: params (i, j, h, w) to be passed to crop for random crop.

VideoRandomHorizontalFlip

class mindvideo.data.transforms.VideoRandomHorizontalFlip(prob=0.5):

Flip every frame of the video with a given probability.

  • base: trans.PyTensorOperation

Parameters:

  • prob (float): probability of the image being flipped. Default: 0.5.

Return:

None

VideoReOrder

class mindvideo.data.transforms.VideoReOrder(new_order):

Rearrange the order of dims of data.

  • base: trans.PyTensorOperation

Parameters:

  • new_order(tuple), new_order of output.

Return:

None

VideoRescale

class mindvideo.data.transforms.VideoRescale(rescale=1 / 255.0, shift=0.0):

Rescale the input video frames with the given rescale and shift. This operator will rescale the input video with: output = image * rescale + shift.

  • base: trans.PyTensorOperation

Parameters:

  • rescale (float): Rescale factor.

  • shift (float, str): Shift factor, if shift is a string, it should be the path to a .npy file with shift data in it.

Return:

None

VideoReshape

class mindvideo.data.transforms.VideoReshape(shape):

Reshape data.

  • base: trans.PyTensorOperation

Parameters:

  • shape(tuple), shape of output.

Return:

None

VideoResize

class mindvideo.data.transforms.VideoResize(size, interpolation=”bilinear”):

Resize the given video sequences (t, h, w, c) at the given size.

  • base: trans.PyTensorOperation

Parameters:

  • size(Union[tuple[int], int]): Desired output size after resize.

  • interpolation (str): TO DO. Default: “bilinear”.

Return:

None

VideoShortEdgeResize

class mindvideo.data.transforms.VideoShortEdgeResize(new_order=(3, 0, 1, 2))):

Resize the given video sequences (t, h, w, x, c) at the given size. And make sure the smallest dimension in (h, w) is 256 pixels.

  • base: trans.PyTensorOperation

Parameters:

  • new_order(tuple), new_order of output.

Return:

None

VideoToTensor

class mindvideo.data.transforms.VideoToTensor(order=(3, 0, 1, 2)):

Convert the input video frames in type numpy.ndarray of shape (T, H, W, C) in the range [0, 255] to numpy.ndarray of shape (C, T, H, W) in the range [-1.0, 1.0] with the desired dtype.

  • base: trans.PyTensorOperation

Parameters:

  • new_order(tuple), new_order of output.

Return:

None

RandomHorizontalFlip

class mindvideo.data.transforms.RandomHorizontalFlip(p=0.5):

Flip every frame of the video ,bboxes and masks with a given probability.

  • base: trans.PyTensorOperation

Parameters:

  • size(int): Desired output size after resize.

  • interpolation (str): TO DO. Default: “bilinear”.

Return:

None

def mindvideo.data.transforms.RandomHorizontalFlip.hflip(clip, boxes, masks, resize_shape, label, valid):

flip img, boxes and masks

ResizeShape

class mindvideo.data.transforms.ResizeShape(size):

Resize img and boxes with at the given size.

  • base: trans.PyTensorOperation

Parameters:

  • size(Union[tuple[int]): Desired output size after resize.

Return:

None

RandomResize

class mindvideo.data.transforms.RandomResize(sizes, max_size=None):

Resize img and boxes with at the given size.

  • base: trans.PyTensorOperation

Parameters:

  • size(Union[tuple[int]): Desired output size after resize.

  • max_size(int): Limit the length after resize. Default:None

Return:

None

def mindvideo.data.transforms.RandomResize.resize(clip, boxes, masks, resize_shape, size, label, valid, max_size=None):

resize img and boxes, then save the resize_shape.size can be min_size (scalar) or (w, h) tuple

PhotometricDistort

class mindvideo.data.transforms.PhotometricDistort():

photometric distortion

  • base: trans.PyTensorOperation

Parameters:

None

Return:

None

RandomContrast

class mindvideo.data.transforms.RandomContrast(lower=0.5, upper=1.5):

random contrast on img

  • base: trans.PyTensorOperation

Parameters:

  • lower(float): smallest random value.Default: 0.5

  • upper(float): largest random value. Default: 1.5

Return:

None

ConvertColor

class mindvideo.data.transforms.ConvertColor(current=’BGR’, transform=’HSV’):

Change image color space.

  • base: trans.PyTensorOperation

Parameters:

  • current(str): current color space.Default: ‘BGR’

  • transform(str): largest random value. Default: ‘HSV’

Return:

None

RandomSaturation

class mindvideo.data.transforms.RandomSaturation(lower=0.5, upper=1.5)

Random saturation on the second channel.

  • base: trans.PyTensorOperation

Parameters:

  • lower(float): smallest random value.Default: 0.5

  • upper(float): largest random value. Default: 1.5

Return:

None

RandomHue

class mindvideo.data.transforms.RandomHue(delta=18.0)

Adjust the hue of RGB images by a random factor.

  • base: trans.PyTensorOperation

Parameters:

  • delta(float):The value to randomly increase or decrease the Hue of image.Default:18.0

Return:

None

RandomBrightness

class mindvideo.data.transforms.RandomBrightness(delta=32)

Adjust the brightness of images by a random factor.

  • base: trans.PyTensorOperation

Parameters:

  • delta(float):the value to random adjust_brightness.Default:32

Return:

None

RandomLightingNoise

class mindvideo.data.transforms.RandomLightingNoise()

Randomly transform channels.6 transformation modes are set, one is randomly selected, and the order of the three BGR channels is changed

  • base: trans.PyTensorOperation

Parameters:

None

Return:

None

SwapChannels

class mindvideo.data.transforms.SwapChannels(swaps)

swap channels.

  • base: trans.PyTensorOperation

Parameters:

  • swap: int

Return:

None

Compose

class mindvideo.data.transforms.Compose(transforms)

compose transform

  • base: trans.PyTensorOperation

Parameters:

  • transforms(list): Data transforms.

Return:

None

RandomSizeCrop

class mindvideo.data.transforms.RandomSizeCrop(min_size: int, max_size: int)

random crop img and boxes. save region data.

  • base: trans.PyTensorOperation

Parameters:

  • min_size(int): the min size of cropped image.

  • max_size(int): the max size of cropped image.

Return:

None

Normalize

class mindvideo.data.transforms.Normalize(mean, std)

normalize on img and boxes.

  • base: trans.PyTensorOperation

Parameters:

  • tmean(list):mean for normalize.

  • std(list):standard deviation for narmalize.

Return:

None

RescaleShape

class mindvideo.data.transforms.RescaleShape(h, w)

resize img,boxes and masks.save resize_shape.

  • base: trans.PyTensorOperation

Parameters:

  • h(int): height of resize shape.

  • w(int): width of resize shape.

Return:

None