o Qe%Sã@sÊddlmZddlmZddlmZddlmZddlm Z ddl Z ddl m Z mZdd lmZdd l mZdd lmZmZdd lmZgZddd„Z   ddd„Zddd„Zddd„Zddd„ZdS)é)Úget_cudnn_version)ÚVariable)Ú LayerHelper)Úcheck_variable_and_dtype)Ú dygraph_utilséN)Ú_C_opsÚ _legacy_C_ops)Úis_compiled_with_rocm)Úin_dynamic_mode)Úin_dygraph_modeÚ_in_legacy_dygraph)Ú_non_static_modeTc CsFt|tƒs tdƒ‚tƒ}|dur|dkr|rd}nd}|jddkr$d}tƒr)d}tƒrAt|tƒr7| ¡ ¡n|}t   ||||¡St ƒr\t|tƒrO| ¡ ¡n|}t   |d|d |d |¡St d ƒ}t|d d dgd ƒ| |j¡}d|i} ||dœ} t|tƒrŠ|| d<t|ddgd ƒn|| d<|jd | d|it| ƒdkrdn| d|S)až It generates a grid of (x,y) or (x,y,z) coordinates using the parameters of the affine transformation that correspond to a set of points where the input feature map should be sampled to produce the transformed output feature map. Args: theta (Tensor) - A tensor with shape [N, 2, 3] or [N, 3, 4]. It contains a batch of affine transform parameters. The data type can be float32 or float64. out_shape (Tensor | list | tuple): Type can be a 1-D Tensor, list, or tuple. It is used to represent the shape of the output in an affine transformation, in the format ``[N, C, H, W]`` or ``[N, C, D, H, W]``. When the format is ``[N, C, H, W]``, it represents the batch size, number of channels, height and width. When the format is ``[N, C, D, H, W]``, it represents the batch size, number of channels, depth, height and width. The data type must be int32. align_corners(bool, optional): if True, aligns the centers of the 4 (4D) or 8 (5D) corner pixels of the input and output tensors, and preserves the value of the corner pixels. Default: True name(str, optional): The default value is None. Normally there is no need for user to set this property. For more information, please refer to :ref:`api_guide_Name`. Returns: Tensor, A Tensor with shape [batch_size, H, W, 2] or [batch, D, H, W, 3] while ('D')'H', 'W' are the (depth)height, width of feature map in affine transformation. The data type is the same as `theta`. Examples: .. code-block:: python import paddle import paddle.nn.functional as F # theta shape = [1, 2, 3] theta = paddle.to_tensor([[[-0.7, -0.4, 0.3], [ 0.6, 0.5, 1.5]]], dtype="float32") y_t = F.affine_grid( theta, [1, 2, 3, 3], align_corners=False) print(y_t) #[[[[ 1.0333333 0.76666665] # [ 0.76666665 1.0999999 ] # [ 0.5 1.4333333 ]] # # [[ 0.5666667 1.1666666 ] # [ 0.3 1.5 ] # [ 0.03333333 1.8333334 ]] # # [[ 0.10000002 1.5666667 ] # [-0.16666666 1.9000001 ] # [-0.43333334 2.2333333 ]]]] zThe theta should be a Tensor.NipTFérZ output_shapeÚ align_cornersÚ use_cudnnÚ affine_gridÚthetaÚfloat32Úfloat64ÚTheta)rrZ OutputShapeÚ out_shapeZint32ÚOutputr©ÚtypeÚinputsÚoutputsÚattrs)Ú isinstancerÚ ValueErrorrÚshaper r ÚnumpyÚtolistrrr r rrÚ"create_variable_for_type_inferenceÚdtypeÚ append_opÚlen) rrrÚnameÚ cudnn_versionrZ _out_shapeÚhelperÚoutÚiptsr©r,úKD:\Projects\ConvertPro\env\Lib\site-packages\paddle/nn/functional/vision.pyrsf .ÿÿýÿýù  ÿ    ÿürÚbilinearÚzeroscCsfddg}gd¢}||vrtd ||¡ƒ‚||vr td ||¡ƒ‚t|tƒs,td |¡ƒ‚tƒ}d} tƒsJ|durJ|rJ|dkrJ|d krJd } d|_d|_t|jƒd krSd} t ƒr_t   |||||¡St ƒrzd |d |d|d| f} t tdƒ||g| ¢RŽ} | Stditƒ¤Ž} t|dddgdƒt|dddgdƒ||dœ} |||| dœ} |  |j¡} | jd| | d| id| S)a… Sample input X by using bilinear interpolation or nearest interpolation based on flow field grid, which is usually generated by :code:`affine_grid` . When the input X is 4-D Tensor, the grid of shape [N, H, W, 2] is the concatenation of (x, y) coordinates with shape [N, H, W] each, where x is indexing the 4th dimension (in width dimension) of input data x and y is indexing the 3rd dimension (in height dimension), finally results is the bilinear interpolation or nearest value of 4 nearest corner points. The output tensor shape will be [N, C, H, W]. When the input X is 5-D Tensor, the grid of shape [N, D, H, W, 3] is the concatenation of (x, y, z) coordinates with shape [N, D, H, W] each, where x is indexing the 5th dimension (in width dimension) of input data x, y is indexing the 4th dimension (in height dimension) and z is indexing the 3rd dimension (in depth dimension) finally results is the bilinear interpolation or nearest value of 8 nearest cornerpoints. The output tensor shape will be [N, C, D, H, W]. Step 1: Get (x, y) grid coordinates and scale to [0, H-1/W-1]. .. code-block:: text grid_x = 0.5 * (grid[:, :, :, 0] + 1) * (W - 1) grid_y = 0.5 * (grid[:, :, :, 1] + 1) * (H - 1) Step 2: Indices input data X with grid (x, y) in each [H, W] area, and bilinear interpolate point value by 4 nearest points or nearest interpolate point value by nearest point. .. code-block:: text wn ------- y_n ------- en | | | | d_n | | | | x_w --d_w-- grid--d_e-- x_e | | | | d_s | | | | ws ------- y_s ------- wn For bilinear interpolation: x_w = floor(x) // west side x coord x_e = x_w + 1 // east side x coord y_n = floor(y) // north side y coord y_s = y_s + 1 // south side y coord d_w = grid_x - x_w // distance to west side d_e = x_e - grid_x // distance to east side d_n = grid_y - y_n // distance to north side d_s = y_s - grid_y // distance to south side wn = X[:, :, y_n, x_w] // north-west point value en = X[:, :, y_n, x_e] // north-east point value ws = X[:, :, y_s, x_w] // south-east point value es = X[:, :, y_s, x_w] // north-east point value output = wn * d_e * d_s + en * d_w * d_s + ws * d_e * d_n + es * d_w * d_n Args: x(Tensor): The input tensor, which is a 4-d tensor with shape [N, C, H, W] or a 5-d tensor with shape [N, C, D, H, W], N is the batch size, C is the channel number, D, H and W is the feature depth, height and width. The data type is float32 or float64. grid(Tensor): Input grid tensor, which is a 4-d tensor with shape [N, grid_H, grid_W, 2] or a 5-d tensor with shape [N, grid_D, grid_H, grid_W, 3]. The data type is float32 or float64. mode(str, optional): The interpolation method which can be 'bilinear' or 'nearest'. Default: 'bilinear'. padding_mode(str, optional) The padding method used when source index is out of input images. It can be 'zeros', 'reflection' and 'border'. Default: zeros. align_corners(bool, optional): If `align_corners` is true, it will projects -1 and 1 to the centers of the corner pixels. Otherwise, it will projects -1 and 1 to the image edges. name(str, optional): For detailed information, please refer to :ref:`api_guide_Name`. Usually name is no need to set and None by default. Returns: Tensor, The shape of output is [N, C, grid_H, grid_W] or [N, C, grid_D, grid_H, grid_W] in which `grid_D` is the depth of grid, `grid_H` is the height of grid and `grid_W` is the width of grid. The data type is same as input tensor. Examples: .. code-block:: python import paddle import paddle.nn.functional as F # x shape=[1, 1, 3, 3] x = paddle.to_tensor([[[[-0.6, 0.8, -0.5], [-0.5, 0.2, 1.2], [ 1.4, 0.3, -0.2]]]],dtype='float64') # grid shape = [1, 3, 4, 2] grid = paddle.to_tensor([[[[ 0.2, 0.3], [-0.4, -0.3], [-0.9, 0.3], [-0.9, -0.6]], [[ 0.4, 0.1], [ 0.9, -0.8], [ 0.4, 0.5], [ 0.5, -0.2]], [[ 0.1, -0.8], [-0.3, -1. ], [ 0.7, 0.4], [ 0.2, 0.8]]]],dtype='float64') y_t = F.grid_sample( x, grid, mode='bilinear', padding_mode='border', align_corners=True) print(y_t) # output shape = [1, 1, 3, 4] # [[[[ 0.34 0.016 0.086 -0.448] # [ 0.55 -0.076 0.35 0.59 ] # [ 0.596 0.38 0.52 0.24 ]]]] r.Znearest)r/Z reflectionÚborderz=The mode of grid sample function should be in {}, but got: {}zEThe padding mode of grid sample function should be in {}, but got: {}z-The align corners should be bool, but got: {}FNr/TéÚmodeÚ padding_moderrZ grid_samplerÚ grid_sampleÚxrrÚgrid)ÚXZGrid)r2r3rrr)rrrr)r4)rÚformatrÚboolrr Z stop_gradientr&r r rr4r Úgetattrr rÚlocalsrr#r$r%)r5r6r2r3rr'Z_modesZ_padding_modesr(rrr*r)r+r,r,r-r4Šsˆ ÿÿÿÿ ÿÿÿÿø í ÿ ü ür4ÚNCHWcCs¨t|tƒs tdƒ‚|dvrtd |¡ƒ‚tƒrt |||¡Stƒr*t  |d|d|¡St dit ƒ¤Ž}t |ddd gdƒ|j |jd }|jdd |id |i||d œd|S)a  This API implements pixel shuffle operation. See more details in :ref:`api_nn_vision_PixelShuffle` . Parameters: x(Tensor): 4-D tensor, the data type should be float32 or float64. upscale_factor(int): factor to increase spatial resolution. data_format (str, optional): The data format of the input and output data. An optional string from: "NCHW", "NHWC". The default is "NCHW". When it is "NCHW", the data is stored in the order of: [batch_size, input_channels, input_height, input_width]. name (str, optional): The default value is None. Normally there is no need for user to set this property. Returns: Out(tensor): Reshaped tensor according to the new dimension. Examples: .. code-block:: python import paddle import paddle.nn.functional as F x = paddle.randn(shape=[2,9,4,4]) out_var = F.pixel_shuffle(x, 3) out = out_var.numpy() print(out.shape) # (2, 1, 12, 12) zupscale factor must be int type©r<ZNHWCúOAttr(data_format) should be 'NCHW' or 'NHWC'.But recevie Attr(data_format): {} Úupscale_factorÚ data_formatÚ pixel_shuffler5rr©r$r7ÚOut)r?r@rN)rA)rÚintÚ TypeErrorrr8r rrAr r rr;rr#r$r%)r5r?r@r'r)r*r,r,r-rA^s. þ ÿürAcCóÂt|jƒdkrtd |j¡ƒ‚t|tƒstdƒ‚|dkr tdƒ‚|dvr+td |¡ƒ‚tƒr7t  |d|d |¡St dit ƒ¤Ž}t |d d d gd ƒ|j |jd}|jd d|id|i||dœd|S)a4 This API implements pixel unshuffle operation. See more details in :ref:`api_nn_vision_PixelUnshuffle` . Parameters: x (Tensor): 4-D tensor, the data type should be float32 or float64. downscale_factor (int): Factor to decrease spatial resolution. data_format (str, optional): The data format of the input and output data. An optional string of NCHW or NHWC. The default is NCHW. When it is NCHW, the data is stored in the order of [batch_size, input_channels, input_height, input_width]. name (str, optional): Name for the operation (optional, default is None). Normally there is no need for user to set this property. For more information, please refer to :ref:`api_guide_Name`. Returns: Out (Tensor): Reshaped tensor according to the new dimension. Examples: .. code-block:: python import paddle import paddle.nn.functional as F x = paddle.randn([2, 1, 12, 12]) out = F.pixel_unshuffle(x, 3) print(out.shape) # [2, 9, 4, 4] éú@Input x should be 4D tensor, but received x with the shape of {}z!Downscale factor must be int typerz!Downscale factor must be positiver=r>Údownscale_factorr@Úpixel_unshuffler5rrrBr7rC)rIr@rN)rJ)r&r rr8rrDrErr rJrr;rr#r$r%)r5rIr@r'r)r*r,r,r-rJ•s>ÿÿ þ ÿþü rJcCrF)a§ This API implements channel shuffle operation. See more details in :ref:`api_nn_vision_ChannelShuffle` . Parameters: x (Tensor): 4-D tensor, the data type should be float32 or float64. groups (int): Number of groups to divide channels in. data_format (str): The data format of the input and output data. An optional string of NCHW or NHWC. The default is NCHW. When it is NCHW, the data is stored in the order of [batch_size, input_channels, input_height, input_width]. name (str, optional): Name for the operation (optional, default is None). Normally there is no need for user to set this property. For more information, please refer to :ref:`api_guide_Name`. Returns: Out (Tensor): Rearranged tensor keeping the original tensor shape. Examples: .. code-block:: python import paddle import paddle.nn.functional as F x = paddle.arange(0, 0.6, 0.1, 'float32') x = paddle.reshape(x, [1, 6, 1, 1]) # [[[[0. ]], # [[0.10000000]], # [[0.20000000]], # [[0.30000001]], # [[0.40000001]], # [[0.50000000]]]] y = F.channel_shuffle(x, 3) # [[[[0. ]], # [[0.20000000]], # [[0.40000001]], # [[0.10000000]], # [[0.30000001]], # [[0.50000000]]]] rGrHzgroups must be int typerzgroups must be positiver=r>Úgroupsr@Úchannel_shuffler5rrrBr7rC)rKr@rN)rL)r&r rr8rrDrErr rLrr;rr#r$r%)r5rKr@r'r)r*r,r,r-rLÔs:#ÿÿ þ ÿürL)TN)r.r/TN)r<N)ZdevicerZstaticrZfluid.layer_helperrZfluid.data_feederrZfluidrr!ÚnpZpaddlerr r r Zpaddle.fluid.frameworkr r Zpaddle.frameworkrÚ__all__rr4rArJrLr,r,r,r-Ús,         o ú U 7?