symbol.image

Image Symbol API of MXNet.

Functions

adjust_lighting([data, alpha, name, attr, out])	Adjust the lighting level of the input.
crop([data, x, y, width, height, name, …])	Crop an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size.
flip_left_right([data, name, attr, out])	Defined in /work/mxnet/src/operator/image/image_random.cc:L199
flip_top_bottom([data, name, attr, out])	Defined in /work/mxnet/src/operator/image/image_random.cc:L211
normalize([data, mean, std, name, attr, out])	Normalize an tensor of shape (C x H x W) or (N x C x H x W) with mean and standard deviation.
random_brightness([data, min_factor, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L223
random_color_jitter([data, brightness, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L251
random_contrast([data, min_factor, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L230
random_crop([data, xrange, yrange, width, …])	Randomly crop an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size.
random_flip_left_right([data, p, name, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L205
random_flip_top_bottom([data, p, name, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L217
random_hue([data, min_factor, max_factor, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L244
random_lighting([data, alpha_std, name, …])	Randomly add PCA noise.
random_resized_crop([data, xrange, yrange, …])	Randomly crop an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size.
random_saturation([data, min_factor, …])	Defined in /work/mxnet/src/operator/image/image_random.cc:L237
resize([data, size, keep_ratio, interp, …])	Resize an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size.
to_tensor([data, name, attr, out])	Converts an image NDArray of shape (H x W x C) or (N x H x W x C) with values in the range [0, 255] to a tensor NDArray of shape (C x H x W) or (N x C x H x W) with values in the range [0, 1].

adjust_lighting(data=None, alpha=_Null, name=None, attr=None, out=None, **kwargs)

Adjust the lighting level of the input. Follow the AlexNet style.

Defined in /work/mxnet/src/operator/image/image_random.cc:L259

Parameters

• data (Symbol) – The input.

• alpha (tuple of <float>, required) – The lighting alphas for the R, G, B channels.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

crop(data=None, x=_Null, y=_Null, width=_Null, height=_Null, name=None, attr=None, out=None, **kwargs)

Crop an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size. Example: .. code-block:: python

image = mx.nd.random.uniform(0, 255, (4, 2, 3)).astype(dtype=np.uint8) mx.nd.image.crop(image, 1, 1, 2, 2).shape # (2, 2, 3) image = mx.nd.random.uniform(0, 255, (2, 4, 2, 3)).astype(dtype=np.uint8) mx.nd.image.crop(image, 1, 1, 2, 2) # (2, 2, 2, 3)

Defined in /work/mxnet/src/operator/image/crop.cc:L49

Parameters

• data (Symbol) – The input.

• x (int, required) – Left boundary of the cropping area.

• y (int, required) – Top boundary of the cropping area.

• width (int, required) – Width of the cropping area.

• height (int, required) – Height of the cropping area.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

flip_left_right(data=None, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L199

Parameters

• data (Symbol) – The input.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

flip_top_bottom(data=None, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L211

Parameters

• data (Symbol) – The input.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

normalize(data=None, mean=_Null, std=_Null, name=None, attr=None, out=None, **kwargs)

Normalize an tensor of shape (C x H x W) or (N x C x H x W) with mean and standard deviation.

Given mean (m1, …, mn) and std (s:sub:`1, …, s<sub>n</sub>)` for n channels, this transform normalizes each channel of the input tensor with:

\[ \begin{align}\begin{aligned} output[i] = (input[i] - m\ :sub:`i`\ ) / s\ :sub:`i`\\If mean or std is scalar, the same value will be applied to all channels.\\Default value for mean is 0.0 and stand deviation is 1.0.\end{aligned}\end{align} \]

Example:

image = mx.nd.random.uniform(0, 1, (3, 4, 2))
normalize(image, mean=(0, 1, 2), std=(3, 2, 1))
    [[[ 0.18293785  0.19761486]
      [ 0.23839645  0.28142193]
      [ 0.20092112  0.28598186]
      [ 0.18162774  0.28241724]]
     [[-0.2881726  -0.18821815]
      [-0.17705294 -0.30780914]
      [-0.2812064  -0.3512327 ]
      [-0.05411351 -0.4716435 ]]
     [[-1.0363373  -1.7273437 ]
      [-1.6165586  -1.5223348 ]
      [-1.208275   -1.1878313 ]
      [-1.4711051  -1.5200229 ]]]
    <NDArray 3x4x2 @cpu(0)>

image = mx.nd.random.uniform(0, 1, (2, 3, 4, 2))
normalize(image, mean=(0, 1, 2), std=(3, 2, 1))
    [[[[ 0.18934818  0.13092826]
       [ 0.3085322   0.27869293]
       [ 0.02367868  0.11246539]
       [ 0.0290431   0.2160573 ]]
      [[-0.4898908  -0.31587923]
       [-0.08369008 -0.02142242]
       [-0.11092162 -0.42982462]
       [-0.06499392 -0.06495637]]
      [[-1.0213816  -1.526392  ]
       [-1.2008414  -1.1990893 ]
       [-1.5385206  -1.4795225 ]
       [-1.2194707  -1.3211205 ]]]
     [[[ 0.03942481  0.24021089]
       [ 0.21330701  0.1940066 ]
       [ 0.04778443  0.17912441]
       [ 0.31488964  0.25287187]]
      [[-0.23907584 -0.4470462 ]
       [-0.29266903 -0.2631998 ]
       [-0.3677222  -0.40683383]
       [-0.11288315 -0.13154092]]
      [[-1.5438497  -1.7834496 ]
       [-1.431566   -1.8647819 ]
       [-1.9812102  -1.675859  ]
       [-1.3823645  -1.8503251 ]]]]
    <NDArray 2x3x4x2 @cpu(0)>

Defined in /work/mxnet/src/operator/image/image_random.cc:L171

Parameters

• data (Symbol) – Input ndarray

• mean (tuple of <float>, optional, default=[0,0,0,0]) – Sequence of means for each channel. Default value is 0.

• std (tuple of <float>, optional, default=[1,1,1,1]) – Sequence of standard deviations for each channel. Default value is 1.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_brightness(data=None, min_factor=_Null, max_factor=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L223

Parameters

• data (Symbol) – The input.

• min_factor (float, required) – Minimum factor.

• max_factor (float, required) – Maximum factor.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_color_jitter(data=None, brightness=_Null, contrast=_Null, saturation=_Null, hue=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L251

Parameters

• data (Symbol) – The input.

• brightness (float, required) – How much to jitter brightness.

• contrast (float, required) – How much to jitter contrast.

• saturation (float, required) – How much to jitter saturation.

• hue (float, required) – How much to jitter hue.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_contrast(data=None, min_factor=_Null, max_factor=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L230

Parameters

• data (Symbol) – The input.

• min_factor (float, required) – Minimum factor.

• max_factor (float, required) – Maximum factor.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_crop(data=None, xrange=_Null, yrange=_Null, width=_Null, height=_Null, interp=_Null, name=None, attr=None, out=None, **kwargs)

Randomly crop an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size. Upsample result if src is smaller than size. Example:

im = mx.nd.array(cv2.imread("flower.jpg"))
cropped_im, rect  = mx.nd.image.random_crop(im, (100, 100))

Defined in /work/mxnet/src/operator/image/crop.cc:L77

Parameters

• data (Symbol) – The input.

• xrange (tuple of <float>, optional, default=[0,1]) – Left boundaries of the cropping area.

• yrange (tuple of <float>, optional, default=[0,1]) – Top boundaries of the cropping area.

• width (int, required) – The target image width

• height (int, required) – The target image height.

• interp (int, optional, default='1') – Interpolation method for resizing. By default uses bilinear interpolationOptions are INTER_NEAREST - a nearest-neighbor interpolationINTER_LINEAR - a bilinear interpolationINTER_AREA - resampling using pixel area relationINTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhoodINTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhoodNote that the GPU version only support bilinear interpolation(1)

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_flip_left_right(data=None, p=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L205

Parameters

• data (Symbol) – The input.

• p (float, optional, default=0.5) – The probablity of flipping the image.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_flip_top_bottom(data=None, p=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L217

Parameters

• data (Symbol) – The input.

• p (float, optional, default=0.5) – The probablity of flipping the image.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_hue(data=None, min_factor=_Null, max_factor=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L244

Parameters

• data (Symbol) – The input.

• min_factor (float, required) – Minimum factor.

• max_factor (float, required) – Maximum factor.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_lighting(data=None, alpha_std=_Null, name=None, attr=None, out=None, **kwargs)

Randomly add PCA noise. Follow the AlexNet style.

Defined in /work/mxnet/src/operator/image/image_random.cc:L266

Parameters

• data (Symbol) – The input.

• alpha_std (float, optional, default=0.0500000007) – Level of the lighting noise.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_resized_crop(data=None, xrange=_Null, yrange=_Null, width=_Null, height=_Null, interp=_Null, name=None, attr=None, out=None, **kwargs)

Randomly crop an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size. Randomize area and aspect ratio. Upsample result if src is smaller than size. Example: .. code-block:: python

im = mx.nd.array(cv2.imread(“flower.jpg”)) cropped_im, rect = mx.nd.image.random_resized_crop(im, (100, 100))

Defined in /work/mxnet/src/operator/image/crop.cc:L114

Parameters

• data (Symbol) – The input.

• xrange (tuple of <float>, optional, default=[0,1]) – Left boundaries of the cropping area.

• yrange (tuple of <float>, optional, default=[0,1]) – Top boundaries of the cropping area.

• width (int, required) – The target image width

• height (int, required) – The target image height.

• interp (int, optional, default='1') – Interpolation method for resizing. By default uses bilinear interpolationOptions are INTER_NEAREST - a nearest-neighbor interpolationINTER_LINEAR - a bilinear interpolationINTER_AREA - resampling using pixel area relationINTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhoodINTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhoodNote that the GPU version only support bilinear interpolation(1)

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

random_saturation(data=None, min_factor=_Null, max_factor=_Null, name=None, attr=None, out=None, **kwargs)

Defined in /work/mxnet/src/operator/image/image_random.cc:L237

Parameters

• data (Symbol) – The input.

• min_factor (float, required) – Minimum factor.

• max_factor (float, required) – Maximum factor.

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

resize(data=None, size=_Null, keep_ratio=_Null, interp=_Null, name=None, attr=None, out=None, **kwargs)

Resize an image NDArray of shape (H x W x C) or (N x H x W x C) to the given size. Example:

image = mx.nd.random.uniform(0, 255, (4, 2, 3)).astype(dtype=np.uint8)
mx.nd.image.resize(image, (3, 3))
    [[[124 111 197]
      [158  80 155]
      [193  50 112]]

     [[110 100 113]
      [134 165 148]
      [157 231 182]]

     [[202 176 134]
      [174 191 149]
      [147 207 164]]]
    <NDArray 3x3x3 @cpu(0)>

image = mx.nd.random.uniform(0, 255, (2, 4, 2, 3)).astype(dtype=np.uint8)
mx.nd.image.resize(image, (2, 2))
    [[[[ 59 133  80]
       [187 114 153]]

      [[ 38 142  39]
       [207 131 124]]]


      [[[117 125 136]
       [191 166 150]]

      [[129  63 113]
       [182 109  48]]]]
    <NDArray 2x2x2x3 @cpu(0)>

Defined in /work/mxnet/src/operator/image/resize.cc:L73

Parameters

• data (Symbol) – The input.

• size (Shape(tuple), optional, default=[]) – Size of new image. Could be (width, height) or (size)

• keep_ratio (boolean, optional, default=0) – Whether to resize the short edge or both edges to size, if size is give as an integer.

• interp (int, optional, default='1') – Interpolation method for resizing. By default uses bilinear interpolationOptions are INTER_NEAREST - a nearest-neighbor interpolationINTER_LINEAR - a bilinear interpolationINTER_AREA - resampling using pixel area relationINTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhoodINTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhoodNote that the GPU version only support bilinear interpolation(1)

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

to_tensor(data=None, name=None, attr=None, out=None, **kwargs)

Converts an image NDArray of shape (H x W x C) or (N x H x W x C) with values in the range [0, 255] to a tensor NDArray of shape (C x H x W) or (N x C x H x W) with values in the range [0, 1].

Examples

>>> image = mx.nd.random.uniform(0, 255, (4, 2, 3)).astype(dtype=np.uint8)
>>> to_tensor(image)
[[[ 0.85490197  0.72156864]
  [ 0.09019608  0.74117649]
  [ 0.61960787  0.92941177]
  [ 0.96470588  0.1882353 ]]
 [[ 0.6156863   0.73725492]
  [ 0.46666667  0.98039216]
  [ 0.44705883  0.45490196]
  [ 0.01960784  0.8509804 ]]
 [[ 0.39607844  0.03137255]
  [ 0.72156864  0.52941179]
  [ 0.16470589  0.7647059 ]
  [ 0.05490196  0.70588237]]]
<NDArray 3x4x2 @cpu(0)>

>>> image = mx.nd.random.uniform(0, 255, (2, 4, 2, 3)).astype(dtype=np.uint8)
>>> to_tensor(image)
[[[[0.11764706 0.5803922 ]
   [0.9411765  0.10588235]
   [0.2627451  0.73333335]
   [0.5647059  0.32156864]]
  [[0.7176471  0.14117648]
   [0.75686276 0.4117647 ]
   [0.18431373 0.45490196]
   [0.13333334 0.6156863 ]]
  [[0.6392157  0.5372549 ]
   [0.52156866 0.47058824]
   [0.77254903 0.21568628]
   [0.01568628 0.14901961]]]
 [[[0.6117647  0.38431373]
   [0.6784314  0.6117647 ]
   [0.69411767 0.96862745]
   [0.67058825 0.35686275]]
  [[0.21960784 0.9411765 ]
   [0.44705883 0.43529412]
   [0.09803922 0.6666667 ]
   [0.16862746 0.1254902 ]]
  [[0.6156863  0.9019608 ]
   [0.35686275 0.9019608 ]
   [0.05882353 0.6509804 ]
   [0.20784314 0.7490196 ]]]]
<NDArray 2x3x4x2 @cpu(0)>

Defined in /work/mxnet/src/operator/image/image_random.cc:L94

Parameters

• data (Symbol) – Input ndarray

• name (string, optional.) – Name of the resulting symbol.

Returns

The result symbol.

Return type

Symbol

---

Did this page help you?

Yes

No

Thanks for your feedback!

---