Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit e265b746 authored by Xusong Wang's avatar Xusong Wang Committed by android-build-merger
Browse files

Sync operation docs between NeuralNetworks.h and HAL file 

am: 66baf741

Change-Id: I2473a12afe7fdf86cf7a2d567bf659424c26601e
parents 2688f8fb 66baf741

current.txt

+1 −1
Original line number Diff line number Diff line
@@ -450,7 +450,7 @@ b83317b66721241887d2770b5ae95fd5af1e77c5daa7530ecb08fae8892f2b43 android.hardwar 
92714960d1a53fc2ec557302b41c7cc93d2636d8364a44bd0f85be0c92927ff8 android.hardware.neuralnetworks@1.2::IExecutionCallback

36e1064c869965dee533c537cefbe87e54db8bd8cd45be7e0e93e00e8a43863a android.hardware.neuralnetworks@1.2::IPreparedModel

e1c734d1545e1a4ae749ff1dd9704a8e594c59aea7c8363159dc258e93e0df3b android.hardware.neuralnetworks@1.2::IPreparedModelCallback

d734c2441b602da240fa0e9afe3b612cdc9f3ae9c1db13216f957861d0673c5e android.hardware.neuralnetworks@1.2::types

209a5ee694b94328afb2af2768f1fe6a69148e2cbb85ec3c340a36eed818c697 android.hardware.neuralnetworks@1.2::types

cf7a4ba516a638f9b82a249c91fb603042c2d9ca43fd5aad9cf6c0401ed2a5d7 android.hardware.nfc@1.2::INfc

abf98c2ae08bf765db54edc8068e36d52eb558cff6706b6fd7c18c65a1f3fc18 android.hardware.nfc@1.2::types

4cb252dc6372a874aef666b92a6e9529915aa187521a700f0789065c3c702ead android.hardware.power.stats@1.0::IPowerStats

neuralnetworks/1.2/types.hal

+138 −40
Original line number Diff line number Diff line
@@ -187,6 +187,10 @@ enum OperationType : int32_t { 
     *     input2.dimension = {5, 4, 3, 1}

     *     output.dimension = {5, 4, 3, 2}

     *

     * Since API level 29, generic zero-sized input tensor is supported. Zero

     * dimension is only compatible with 0 or 1. The size of the output

     * dimension is zero if either of corresponding input dimension is zero.

     *

     * Supported tensor {@link OperandType}:

     * * {@link OperandType::TENSOR_FLOAT16} (since API level 29)

     * * {@link OperandType::TENSOR_FLOAT32}
@@ -236,7 +240,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (explicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the padding on

     *      the left, in the ‘width’ dimension.

     * * 2: An {@link OperandType::INT32} scalar, specifying the padding on
@@ -262,7 +267,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (implicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the implicit

     *      padding scheme, has to be one of the

     *      following values: {0 (NONE), 1 (SAME), 2 (VALID)}.
@@ -309,6 +315,7 @@ enum OperationType : int32_t { 
     *            Before API level 29, all input tensors of

     *            {@link OperandType::TENSOR_QUANT8_ASYMM}

     *            must have the same scale and zeroPoint as the output tensor.

     *            Since API level 29, zero-sized tensors are supported.

     * * n: An {@link OperandType::INT32} scalar, specifying the

     *      concatenation axis.

     *
@@ -366,7 +373,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (explicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth_in],

     *      specifying the input.

     *      specifying the input. Since API level 29, zero batches is supported

     *      for this tensor.

     * * 1: A 4-D tensor, of shape

     *      [depth_out, filter_height, filter_width, depth_in], specifying the

     *      filter. For tensor of type
@@ -413,7 +421,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (implicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth_in],

     *      specifying the input.

     *      specifying the input. Since API level 29, zero batches is supported

     *      for this tensor.

     * * 1: A 4-D tensor, of shape

     *      [depth_out, filter_height, filter_width, depth_in], specifying the

     *      filter. For tensor of type
@@ -675,7 +684,7 @@ enum OperationType : int32_t { 
     * Supported tensor rank: up to 4

     *

     * Inputs:

     * * 0: A tensor.

     * * 0: A tensor. Since API level 29, this tensor may be zero-sized.

     *

     * Outputs:

     * * 0: A tensor with the same shape as input0.
@@ -768,7 +777,8 @@ enum OperationType : int32_t { 
     *      [batch_size, input_size], where "input_size" corresponds to the

     *      number of inputs to the layer, matching the second dimension of

     *      weights, and "batch_size" is calculated by dividing the number of

     *      elements by "input_size".

     *      elements by "input_size". Since API level 29, zero batch_size is

     *      supported for this tensor.

     * * 1: A 2-D tensor, specifying the weights, of shape

     *      [num_units, input_size], where "num_units" corresponds to the number

     *      of output nodes.
@@ -864,6 +874,7 @@ enum OperationType : int32_t { 
     * Supported tensor {@link OperandType}:

     * * {@link OperandType::TENSOR_FLOAT16} (since API level 29)

     * * {@link OperandType::TENSOR_FLOAT32}

     * * {@link OperandType::TENSOR_QUANT8_ASYMM} (since API level 29)

     *

     * Supported tensor rank: up to 4

     * Tensors with rank less than 4 are only supported since API level 29.
@@ -878,6 +889,8 @@ enum OperationType : int32_t { 
     *

     * Outputs:

     * * 0: A tensor of the same {@link OperandType} and same shape as input0.

     *      For {@link OperandType::TENSOR_QUANT8_ASYMM},

     *      the scale must be 1.f / 128 and the zeroPoint must be 128.

     *

     * Available since API level 27.

     */
@@ -908,7 +921,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (explicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the padding on

     *      the left, in the ‘width’ dimension.

     * * 2: An {@link OperandType::INT32} scalar, specifying the padding on
@@ -934,7 +948,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (implicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the implicit

     *      padding scheme, has to be one of the

     *      following values: {0 (NONE), 1 (SAME), 2 (VALID)}.
@@ -1024,7 +1039,8 @@ enum OperationType : int32_t { 
     * Supported tensor rank: up to 4.

     *

     * Inputs:

     * * 0: A tensor, specifying the input.

     * * 0: A tensor, specifying the input. Since API level 29, this tensor may

     *      be zero-sized.

     *

     * Outputs:

     * * 0: The output tensor of same shape as input0.
@@ -1336,7 +1352,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (explicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the padding on

     *      the left, in the ‘width’ dimension.

     * * 2: An {@link OperandType::INT32} scalar, specifying the padding on
@@ -1362,7 +1379,8 @@ enum OperationType : int32_t { 
     *

     * Inputs (implicit padding):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the implicit

     *      padding scheme, has to be one of the

     *      following values: {0 (NONE), 1 (SAME), 2 (VALID)}.
@@ -1409,6 +1427,10 @@ enum OperationType : int32_t { 
     * * {@link OperandType::TENSOR_FLOAT32}

     * * {@link OperandType::TENSOR_QUANT8_ASYMM}

     *

     * Since API level 29, generic zero-sized input tensor is supported. Zero

     * dimension is only compatible with 0 or 1. The size of the output

     * dimension is zero if either of corresponding input dimension is zero.

     *

     * Supported tensor rank: up to 4

     *

     * Inputs:
@@ -1444,7 +1466,8 @@ enum OperationType : int32_t { 
     * Supported tensor rank: up to 4.

     *

     * Inputs:

     * * 0: A tensor, specifying the input.

     * * 0: A tensor, specifying the input. Since API level 29, this tensor may

     *      be zero-sized.

     *

     * Outputs:

     * * 0: The output tensor of same shape as input0.
@@ -1468,7 +1491,8 @@ enum OperationType : int32_t { 
     * Supported tensor rank: up to 4.

     *

     * Inputs:

     * * 0: A tensor, specifying the input.

     * * 0: A tensor, specifying the input. Since API level 29, this tensor may

     *      be zero-sized.

     *

     * Outputs:

     * * 0: The output tensor of same shape as input0.
@@ -1492,7 +1516,8 @@ enum OperationType : int32_t { 
     * Supported tensor rank: up to 4.

     *

     * Inputs:

     * * 0: A tensor, specifying the input.

     * * 0: A tensor, specifying the input. Since API level 29, this tensor may

     *      be zero-sized.

     *

     * Outputs:

     * * 0: The output tensor of same shape as input0.
@@ -1544,9 +1569,12 @@ enum OperationType : int32_t { 
     * [batch, height, width, channels]. Alternatively, the data layout could

     * be NCHW, the data storage order of: [batch, channels, height, width].

     *

     * Inputs:

     * Both resizing by shape and resizing by scale are supported.

     *

     * Inputs (resizing by shape):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Since API level 29, zero batches is supported for this

     *      tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the output

     *      height of the output tensor.

     * * 2: An {@link OperandType::INT32} scalar, specifying the output
@@ -1555,6 +1583,24 @@ enum OperationType : int32_t { 
     *      Set to true to specify NCHW data layout for input0 and output0.

     *      Available since API level 29.

     *

     * Inputs (resizing by scale, since API level 29):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input. Zero batches is supported for this tensor.

     * * 1: A scalar, specifying height_scale, the scaling factor of the height

     *      dimension from the input tensor to the output tensor. The output

     *      height is calculated as new_height = floor(height * height_scale).

     *      The scalar must be of {@link OperandType::FLOAT16} if input0 is

     *      of {@link OperandType::TENSOR_FLOAT16} and of

     *      {@link OperandType::FLOAT32} otherwise.

     * * 2: A scalar, specifying width_scale, the scaling factor of the width

     *      dimension from the input tensor to the output tensor. The output

     *      width is calculated as new_width = floor(width * width_scale).

     *      The scalar must be of {@link OperandType::FLOAT16} if input0 is

     *      of {@link OperandType::TENSOR_FLOAT16} and of

     *      {@link OperandType::FLOAT32} otherwise.

     * * 3: An optional {@link OperandType::BOOL} scalar, default to false.

     *      Set to true to specify NCHW data layout for input0 and output0.

     *

     * Outputs:

     * * 0: The output 4-D tensor, of shape

     *      [batches, new_height, new_width, depth].
@@ -1640,7 +1686,8 @@ enum OperationType : int32_t { 
     * Tensors with rank other than 2 or 4 are only supported since API level 29.

     *

     * Inputs:

     * * 0: A 2-D or 4-D tensor, specifying the tensor to be reshaped.

     * * 0: A 2-D or 4-D tensor, specifying the tensor to be reshaped. Since

     *      API level 29, this tensor may be zero-sized.

     * * 1: A scalar, specifying the positive scaling factor for the exponent,

     *      beta. If input0 is of {@link OperandType::TENSOR_FLOAT32} or

     *      {@link OperandType::TENSOR_QUANT8_ASYMM}, the scalar must be of
@@ -1798,7 +1845,8 @@ enum OperationType : int32_t { 
     * Supported tensor rank: up to 4.

     *

     * Inputs:

     * * 0: A tensor, specifying the input.

     * * 0: A tensor, specifying the input. Since API level 29, this tensor may

     *      be zero-sized.

     *

     * Outputs:

     * * 0: The output tensor of same shape as input0.
@@ -1865,6 +1913,10 @@ enum OperationType : int32_t { 
     *     input2.dimension = {5, 4, 3, 1}

     *     output.dimension = {5, 4, 3, 2}

     *

     * Since API level 29, generic zero-sized input tensor is supported. Zero

     * dimension is only compatible with 0 or 1. The size of the output

     * dimension is zero if either of corresponding input dimension is zero.

     *

     * Supported tensor {@link OperandType}:

     * * {@link OperandType::TENSOR_FLOAT16} (since API level 29)

     * * {@link OperandType::TENSOR_FLOAT32}
@@ -2098,6 +2150,10 @@ enum OperationType : int32_t { 
     *     input2.dimension = {5, 4, 3, 1}

     *     output.dimension = {5, 4, 3, 2}

     *

     * Since API level 29, generic zero-sized input tensor is supported. Zero

     * dimension is only compatible with 0 or 1. The size of the output

     * dimension is zero if either of corresponding input dimension is zero.

     *

     * Supported tensor {@link OperandType}:

     * * {@link OperandType::TENSOR_FLOAT16} (since API level 29)

     * * {@link OperandType::TENSOR_FLOAT32}
@@ -2138,6 +2194,7 @@ enum OperationType : int32_t { 
     *

     * Inputs:

     * * 0: An n-D tensor, specifying the tensor to be transposed.

     *      Since API level 29, this tensor may be zero-sized.

     * * 1: An optional 1-D Tensor of {@link OperandType::TENSOR_INT32},

     *      the permutation of the dimensions of the input tensor.

     *
@@ -2234,7 +2291,8 @@ enum OperationType : int32_t { 
     * * 0: A 2-D Tensor of shape [num_rois, 4], specifying the locations of the

     *      bounding box proposals, each line with format [x1, y1, x2, y2].

     *      For tensor of type {@link OperandType::TENSOR_QUANT16_ASYMM},

     *      the zeroPoint must be 0 and the scale must be 0.125.

     *      the zeroPoint must be 0 and the scale must be 0.125. Zero num_rois

     *      is supported for this tensor.

     * * 1: A 2-D Tensor of shape [num_rois, num_classes * 4], specifying the

     *      bounding box delta for each region of interest and each class. The

     *      bounding box deltas are organized in the following order
@@ -2243,10 +2301,12 @@ enum OperationType : int32_t { 
     *      and height, dw and dh is the log-scale relative correction factor

     *      for the width and height. For input0 of type

     *      {@link OperandType::TENSOR_QUANT16_ASYMM}, this tensor should be

     *      of {@link OperandType::TENSOR_QUANT8_ASYMM}.

     *      of {@link OperandType::TENSOR_QUANT8_ASYMM}. Zero num_rois is

     *      supported for this tensor.

     * * 2: An 1-D {@link OperandType::TENSOR_INT32} tensor, of shape

     *      [num_rois], specifying the batch index of each box. Boxes with

     *      the same batch index are grouped together.

     *      the same batch index are grouped together. Zero num_rois is

     *      supported for this tensor.

     * * 3: A 2-D Tensor of shape [batches, 2], specifying the information of

     *      each image in the batch, each line with format

     *      [image_height, image_width].
@@ -2580,10 +2640,17 @@ enum OperationType : int32_t { 
    /**

     * Greedily selects a subset of bounding boxes in descending order of score.

     *

     * This op applies hard NMS algorithm to each class. In each loop of

     * execution, the box with maximum score gets selected, and any boxes with

     * the intersection-over-union (IOU) greater than a threshold are removed

     * from the pending set.

     * This op applies NMS algorithm to each class. In each loop of execution,

     * the box with maximum score gets selected and removed from the pending set.

     * The scores of the rest of boxes are lowered according to the

     * intersection-over-union (IOU) overlapping with the previously selected

     * boxes and a specified NMS kernel method. Any boxes with score less

     * than a threshold are removed from the pending set.

     *

     * Three NMS kernels are supported:

     * * Hard:     score_new = score_old * (1 if IoU < threshold else 0)

     * * Linear:   score_new = score_old * (1 if IoU < threshold else 1 - IoU)

     * * Gaussian: score_new = score_old * exp(- IoU^2 / sigma)

     *

     * Axis-aligned bounding boxes are represented by its upper-left corner

     * coordinate (x1,y1) and lower-right corner coordinate (x2,y2). A valid
@@ -2597,25 +2664,34 @@ enum OperationType : int32_t { 
     * Inputs:

     * * 0: A 2-D Tensor of shape [num_rois, num_classes], specifying the score

     *      of each bounding box proposal. The boxes are grouped by batches in the

     *      first dimension.

     *      first dimension. Zero num_rois is supported for this tensor.

     * * 1: A 2-D Tensor specifying the bounding boxes of shape

     *      [num_rois, num_classes * 4], organized in the order [x1, y1, x2, y2].

     *      The boxes are grouped by batches in the first dimension. The sequential

     *      order of the boxes corresponds with input0. For input0 of type

     *      {@link OperandType::TENSOR_QUANT8_ASYMM}, this tensor should be of

     *      {@link OperandType::TENSOR_QUANT16_ASYMM}, with zeroPoint of 0 and

     *      scale of 0.125.

     *      scale of 0.125. Zero num_rois is supported for this tensor.

     * * 2: A 1-D {@link OperandType::TENSOR_INT32} tensor, of shape

     *      [num_rois], specifying the batch index of each box. Boxes with

     *      the same batch index are grouped together.

     * * 3: An {@link OperandType::FLOAT32} scalar, score_threshold. Boxes

     *      with scores lower than the threshold are filtered before sending

     *      to the NMS algorithm.

     * * 4: An {@link OperandType::FLOAT32} scalar, specifying the IoU

     *      threshold.

     * * 5: An {@link OperandType::INT32} scalar, specifying the maximum

     * * 4: An {@link OperandType::INT32} scalar, specifying the maximum

     *      number of selected bounding boxes for each image. Set to a negative

     *      value for unlimited number of output bounding boxes.

     * * 5: An {@link OperandType::INT32} scalar, specifying the NMS

     *      kernel method, options are 0:hard, 1:linear, 2:gaussian.

     * * 6: An {@link OperandType::FLOAT32} scalar, specifying the IoU

     *      threshold in hard and linear NMS kernel. This field is ignored if

     *      gaussian kernel is selected.

     * * 7: An {@link OperandType::FLOAT32} scalar, specifying the sigma in

     *      gaussian NMS kernel. This field is ignored if gaussian kernel is

     *      not selected.

     * * 8: An {@link OperandType::FLOAT32} scalar, nms_score_threshold.

     *      Boxes with scores lower than the threshold are dropped during the

     *      score updating phase in soft NMS.

     *

     * Outputs:

     * * 0: A 1-D Tensor of the same {@link OperandType} as input0, with shape
@@ -2633,8 +2709,8 @@ enum OperationType : int32_t { 
     *      [num_output_rois], specifying the class of each output box. The

     *      sequential order of the boxes corresponds with output0.

     * * 3: A 1-D {@link OperandType::TENSOR_INT32} tensor, of shape

     *      [num_rois], specifying the batch index of each box. Boxes with

     *      the same batch index are grouped together.

     *      [num_output_rois], specifying the batch index of each box. Boxes

     *      with the same batch index are grouped together.

     *

     * Available since API level 29.

     */
@@ -2970,8 +3046,8 @@ enum OperationType : int32_t { 
     *      For type of {@link OperandType::TENSOR_QUANT16_ASYMM}, the

     *      scale must be 0.125 and the zero point must be 0.

     * * 2: A 1-D {@link OperandType::TENSOR_INT32} tensor, of shape

     *      [num_rois], specifying the batch index of each box. Boxes with

     *      the same batch index are grouped together.

     *      [num_output_rois], specifying the batch index of each box. Boxes

     *      with the same batch index are grouped together.

     *

     * Available since API level 29.

     */
@@ -3637,7 +3713,7 @@ enum OperationType : int32_t { 
     * Supported tensor rank: from 1

     *

     * Inputs:

     * * 0: A tensor.

     * * 0: A tensor, may be zero-sized.

     *

     * Outputs:

     * * 0: The output tensor of same shape as input0, but with
@@ -3969,10 +4045,12 @@ enum OperationType : int32_t { 
     *      the regions of interest, each line with format [x1, y1, x2, y2].

     *      For input0 of type {@link OperandType::TENSOR_QUANT8_ASYMM},

     *      this tensor should be of {@link OperandType::TENSOR_QUANT16_ASYMM},

     *      with zeroPoint of 0 and scale of 0.125.

     *      with zeroPoint of 0 and scale of 0.125. Zero num_rois is

     *      supported for this tensor.

     * * 2: An 1-D {@link OperandType::TENSOR_INT32} tensor, of shape

     *      [num_rois], specifying the batch index of each box. Boxes with

     *      the same batch index are grouped together.

     *      the same batch index are grouped together. Zero num_rois is

     *      supported for this tensor.

     * * 3: An {@link OperandType::INT32} scalar, specifying the output

     *      height of the output tensor.

     * * 4: An {@link OperandType::INT32} scalar, specifying the output
@@ -4137,7 +4215,7 @@ enum OperationType : int32_t { 
     * Supported tensor rank: from 1

     *

     * Inputs:

     * * 0: An n-D tensor to take slice from.

     * * 0: An n-D tensor to take slice from, may be zero-sized.

     * * 1: A 1-D tensor of type {@link OperandType::TENSOR_INT32} specifying

     *      the beginning indices of the slice in each dimension.

     * * 2: A 1-D tensor of type {@link OperandType::TENSOR_INT32} specifying
@@ -4551,9 +4629,11 @@ enum OperationType : int32_t { 
     * [batch, height, width, channels]. Alternatively, the data layout could

     * be NCHW, the data storage order of: [batch, channels, height, width].

     *

     * Inputs:

     * Both resizing by shape and resizing by scale are supported.

     *

     * Inputs (resizing by shape):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input.

     *      the input. Zero batches is supported for this tensor.

     * * 1: An {@link OperandType::INT32} scalar, specifying the output

     *      height of the output tensor.

     * * 2: An {@link OperandType::INT32} scalar, specifying the output
@@ -4561,6 +4641,24 @@ enum OperationType : int32_t { 
     * * 3: An {@link OperandType::BOOL} scalar, default to false.

     *      Set to true to specify NCHW data layout for input0 and output0.

     *

     * Inputs (resizing by scale):

     * * 0: A 4-D tensor, of shape [batches, height, width, depth], specifying

     *      the input. Zero batches is supported for this tensor.

     * * 1: A scalar, specifying height_scale, the scaling factor of the height

     *      dimension from the input tensor to the output tensor. The output

     *      height is calculated as new_height = floor(height * height_scale).

     *      The scalar must be of {@link OperandType::FLOAT16} if input0 is

     *      of {@link OperandType::TENSOR_FLOAT16} and of

     *      {@link OperandType::FLOAT32} otherwise.

     * * 2: A scalar, specifying width_scale, the scaling factor of the width

     *      dimension from the input tensor to the output tensor. The output

     *      width is calculated as new_width = floor(width * width_scale).

     *      The scalar must be of {@link OperandType::FLOAT16} if input0 is

     *      of {@link OperandType::TENSOR_FLOAT16} and of

     *      {@link OperandType::FLOAT32} otherwise.

     * * 3: An {@link OperandType::BOOL} scalar, default to false.

     *      Set to true to specify NCHW data layout for input0 and output0.

     *

     * Outputs:

     * * 0: The output 4-D tensor, of shape

     *      [batches, new_height, new_width, depth].