o .wi} @sXddlmZddlmZddlZddlmZddlmZde de dej d eej e fd ef d d Zd e dee dee dej d eej e fd ef ddZdede d eeeffddZd$dede de de d ef ddZd$dede de d efddZdede d efddZd e dee dee dej d ej d ef ddZdede d e d!e d ef d"d#ZdS)%)Sequence)UnionN)Tensor) functional kernel_sizesigmadtypedevicereturncCsTtjd|dd|dd||d}tt||d d}||jddS)aYCompute 1D gaussian kernel. Args: kernel_size: size of the gaussian kernel sigma: Standard deviation of the gaussian kernel dtype: data type of the output tensor device: device of the output tensor Example: >>> _gaussian(3, 1, torch.float, 'cpu') tensor([[0.2741, 0.4519, 0.2741]]) )startendsteprr rdim)torcharangeexppowsum unsqueeze)rrrr distgaussr`/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/torchmetrics/functional/image/utils.py _gaussian s$rchannelcCsTt|d|d||}t|d|d||}t||}||d|d|dS)aCompute 2D gaussian kernel. Args: channel: number of channels in the image kernel_size: size of the gaussian kernel as a tuple (h, w) sigma: Standard deviation of the gaussian kernel dtype: data type of the output tensor device: device of the output tensor Example: >>> _gaussian_kernel_2d(1, (5,5), (1,1), torch.float, "cpu") tensor([[[[0.0030, 0.0133, 0.0219, 0.0133, 0.0030], [0.0133, 0.0596, 0.0983, 0.0596, 0.0133], [0.0219, 0.0983, 0.1621, 0.0983, 0.0219], [0.0133, 0.0596, 0.0983, 0.0596, 0.0133], [0.0030, 0.0133, 0.0219, 0.0133, 0.0030]]]]) rr )rrmatmultexpand)rrrrr gaussian_kernel_xgaussian_kernel_ykernelrrr_gaussian_kernel_2dsr$inputs window_sizecCsBtjdd|||j|jd}||d}tjd|j|jd}||fS)zConstruct uniform weight and bias for a 2d convolution. Args: inputs: Input image window_size: size of convolutional kernel Return: The weight and bias for 2d convolution r )rr r )ronesrr zeros)r%r& kernel_weight kernel_biasrrr_uniform_weight_bias_conv2d<s r+rpad outer_padc Csl|j|}t||t|ddd|j}t||t|d|||d|j}t|||f|S)a,Apply single-dimension reflection padding to match scipy implementation. Args: inputs: Input image dim: A dimension the image should be padded over pad: Number of pads outer_pad: Number of outer pads Return: Image padded over a single dimension r )shaper index_selectrtor cat)r%rr,r-_maxxyrrr_single_dimension_padMs $,r6cCsdD] }t||||}q|S)zApply reflection padding to the input image. Args: inputs: Input image pad: Number of pads outer_pad: Number of outer pads Return: Padded image )r )r6)r%r,r-rrrr_reflection_pad_2d`s r8csLt|d|dt|\tjfddtjdDddS)zApply uniform filter with a window of a given size over the input image. Args: inputs: Input image window_size: Sliding window used for rmse calculation Return: Image transformed with the uniform input r cs0g|]}tjdd|fdddqS)Nr r)padding)Fconv2dr).0rr%r*r)rr s"z#_uniform_filter..r r)r8r+rr2ranger/)r%r&rr=r_uniform_filterqs  r@c Cst|d|d||}t|d|d||}t|d|d||}t||}t|ddd|d||d|d|d} | |d|d|d|dS)a6Compute 3D gaussian kernel. Args: channel: number of channels in the image kernel_size: size of the gaussian kernel as a tuple (h, w, d) sigma: Standard deviation of the gaussian kernel dtype: data type of the output tensor device: device of the output tensor rr r r.)rrrrmulrrepeatr ) rrrrr r!r"gaussian_kernel_z kernel_xyr#rrr_gaussian_kernel_3ds rEpad_hpad_wpad_dcCstj|||||||fddS)aUReflective padding of 3d input. Args: inputs: tensor to pad, should be a 3D tensor of shape ``[N, C, H, W, D]`` pad_w: amount of padding in the height dimension pad_h: amount of padding in the width dimension pad_d: amount of padding in the depth dimension Returns: padded input tensor reflect)mode)r:r,)r%rFrGrHrrr_reflection_pad_3ds rK)r)collections.abcrtypingrrrtorch.nnrr:intfloatrr strrr$tupler+r6r8r@rErKrrrrsJ   *    "