qiV0dZddlZddlmZmZmZddlmZm Z m Z m Z m Z ddl mZmZmZmZmZmZmZmZmZmZmZmZddlmZddlmZmZmZdd l m!Z!m"Z"ejFe$Z%e!rddl&Z&Gd d ed Z'e"dGddeZ(dgZ)y)z Image processor class for Donut.N)BaseImageProcessor BatchFeature get_size_dict)convert_to_rgbget_resize_output_image_sizepadresizeto_channel_dimension_format) IMAGENET_STANDARD_MEANIMAGENET_STANDARD_STDChannelDimension ImageInputPILImageResamplingget_image_sizeinfer_channel_dimension_formatis_scaled_imagemake_flat_list_of_imagesto_numpy_array valid_imagesvalidate_preprocess_arguments) ImagesKwargs) TensorTypefilter_out_non_signature_kwargslogging)is_vision_availablerequiresc&eZdZUdZeed<eed<y)DonutImageProcessorKwargsaI do_thumbnail (`bool`, *optional*, defaults to `self.do_thumbnail`): Whether to resize the image using thumbnail method. do_align_long_axis (`bool`, *optional*, defaults to `self.do_align_long_axis`): Whether to align the long axis of the image with the long axis of `size` by rotating by 90 degrees. do_thumbnaildo_align_long_axisN)__name__ __module__ __qualname____doc__bool__annotations__b/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/donut/image_processing_donut.pyrr4sr)rF)total)vision)backendsc$eZdZdZdgZeZddejddddddddf de de e e fdzd ed e d e d e d e de e zde de ee zdzde ee zdzddffd Z ddej"de e e fde ezdzde ezdzdej"f dZ d dej"de e e fde de ezdzde ezdzdej"f dZej*ddfdej"de e e fd ede ezdzde ezdzdej"f dZej*ddfdej"de e e fd ede ezdzde ezdzdej"f dZedddddddddddddej2dfdede dzde e e fdzd edzd e dzd e dzd e dzde d e dzde dzde dzde ee zdzde ee zdzde ezdzdedzde ezdzdej:j:f"dZxZS)!DonutImageProcessora Constructs a Donut image processor. Args: do_resize (`bool`, *optional*, defaults to `True`): Whether to resize the image's (height, width) dimensions to the specified `size`. Can be overridden by `do_resize` in the `preprocess` method. size (`dict[str, int]` *optional*, defaults to `{"shortest_edge": 224}`): Size of the image after resizing. The shortest edge of the image is resized to size["shortest_edge"], with the longest edge resized to keep the input aspect ratio. Can be overridden by `size` in the `preprocess` method. resample (`PILImageResampling`, *optional*, defaults to `Resampling.BILINEAR`): Resampling filter to use if resizing the image. Can be overridden by `resample` in the `preprocess` method. do_thumbnail (`bool`, *optional*, defaults to `True`): Whether to resize the image using thumbnail method. do_align_long_axis (`bool`, *optional*, defaults to `False`): Whether to align the long axis of the image with the long axis of `size` by rotating by 90 degrees. do_pad (`bool`, *optional*, defaults to `True`): Whether to pad the image. If `random_padding` is set to `True` in `preprocess`, each image is padded with a random amount of padding on each size, up to the largest image size in the batch. Otherwise, all images are padded to the largest image size in the batch. do_rescale (`bool`, *optional*, defaults to `True`): Whether to rescale the image by the specified scale `rescale_factor`. Can be overridden by `do_rescale` in the `preprocess` method. rescale_factor (`int` or `float`, *optional*, defaults to `1/255`): Scale factor to use if rescaling the image. Can be overridden by `rescale_factor` in the `preprocess` method. do_normalize (`bool`, *optional*, defaults to `True`): Whether to normalize the image. Can be overridden by `do_normalize` in the `preprocess` method. image_mean (`float` or `list[float]`, *optional*, defaults to `IMAGENET_STANDARD_MEAN`): Mean to use if normalizing the image. This is a float or list of floats the length of the number of channels in the image. Can be overridden by the `image_mean` parameter in the `preprocess` method. image_std (`float` or `list[float]`, *optional*, defaults to `IMAGENET_STANDARD_STD`): Image standard deviation. pixel_valuesTNFgp? do_resizesizeresampler r!do_pad do_rescalerescale_factor do_normalize image_mean image_stdreturnc  Nt |di| ||nddd}t|ttfr|ddd}t |}||_||_||_||_ ||_ ||_ ||_ ||_ | |_| | nt|_| | |_yt"|_y)Ni i)heightwidthr()super__init__ isinstancetuplelistrr1r2r3r r!r4r5r6r7r r8r r9)selfr1r2r3r r!r4r5r6r7r8r9kwargs __class__s r*r@zDonutImageProcessor.__init__is "6"'tt-L dUDM *":DT""   ("4 $,((2(>*DZ&/&;AVr)image data_formatinput_data_formatc@t||\}}|d|d}}| t|}|tjk(rd} n$|tjk(rd} nt d|||kr||kDs ||kDr||krt j|d| }|t||| }|S) a Align the long axis of the image to the longest axis of the specified size. Args: image (`np.ndarray`): The image to be aligned. size (`dict[str, int]`): The size `{"height": h, "width": w}` to align the long axis to. data_format (`str` or `ChannelDimension`, *optional*): The data format of the output image. If unset, the same format as the input image is used. input_data_format (`ChannelDimension` or `str`, *optional*): The channel dimension format of the input image. If not provided, it will be inferred. Returns: `np.ndarray`: The aligned image.  channel_dimr<r=)r)rMzUnsupported data format: r)axesinput_channel_dim) rrrLASTFIRST ValueErrornprot90r ) rDrGr2rHrI input_height input_width output_height output_widthrot_axess r*align_long_axisz#DonutImageProcessor.align_long_axiss.%35FW$X! k&*8nd7m|  $ >u E   0 5 5 5H "2"8"8 8H89J8KLM M = ([<-G = ([<-GHHUAH5E  "/{VghE r)random_paddingc6|d|d}}t||\}} || z } ||z } |rItjjd| dz} tjjd| dz} n | dz} | dz} | | z }| | z }| |f| |ff}t ||||S) a Pad the image to the specified size. Args: image (`np.ndarray`): The image to be padded. size (`dict[str, int]`): The size `{"height": h, "width": w}` to pad the image to. random_padding (`bool`, *optional*, defaults to `False`): Whether to use random padding or not. data_format (`str` or `ChannelDimension`, *optional*): The data format of the output image. If unset, the same format as the input image is used. input_data_format (`ChannelDimension` or `str`, *optional*): The channel dimension format of the input image. If not provided, it will be inferred. r<r=rKrrM)lowhighrN)rHrI)rrUrandomrandintr )rDrGr2r]rHrIrYrZrWrX delta_width delta_heightpad_toppad_left pad_bottom pad_rightpaddings r* pad_imagezDonutImageProcessor.pad_images.'+8nd7m| $25FW$X! k"[0 $|3 ii''AL14D'EGyy((Q[1_(EH"a'G"a'H!G+ (* Z(8Y*?@5'{N_``r)c t||\}}|d|d} } t|| } t|| } | |k(r| |k(r|S||kDrt|| z|z } n||kDrt|| z|z } t|f| | f|d||d|S)as Resize the image to make a thumbnail. The image is resized so that no dimension is larger than any corresponding dimension of the specified size. Args: image (`np.ndarray`): The image to be resized. size (`dict[str, int]`): The size `{"height": h, "width": w}` to resize the image to. resample (`PILImageResampling`, *optional*, defaults to `PILImageResampling.BICUBIC`): The resampling filter to use. data_format (`Optional[Union[str, ChannelDimension]]`, *optional*): The data format of the output image. If unset, the same format as the input image is used. input_data_format (`ChannelDimension` or `str`, *optional*): The channel dimension format of the input image. If not provided, it will be inferred. rKr<r=g@)r2r3 reducing_gaprHrI)rminintr ) rDrGr2r3rHrIrErWrXrYrZr<r=s r* thumbnailzDonutImageProcessor.thumbnails2%35FW$X! k&*8nd7m| \=1K. \ !e{&:L + % f,|;zSInvalid image type. Must be of type PIL.Image.Image, numpy.ndarray, or torch.Tensor)r5r6r7r8r9r1r2r3rzIt looks like you are trying to rescale already rescaled images. If the input images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again.)r2rI)rGr2r3rI)rGr2rI)rGr2r]rI)rGscalerI)rGmeanstdrIrPr0)data tensor_type)!r1r2rArBrCrr3r r!r4r5r6r7r8r9rrrTrrrrlogger warning_oncerr\r rorjrescale normalizer r)rDrur1r2r3r r!r4r]r5r6r7r8r9rvrHrIrGr{s r* preprocesszDonutImageProcessor.preprocess>s/R"+!6IDNN 'tTYY dUDM *":DT"'38'3'?|TEVEV 3E3Q/W[WnWn!-4;;#-#9Zt +9+E4K^K^'3'?|TEVEV #-#9Zt !*!6IDNN )&1F#rs s%!)%! 6<vay I  ouvfkd**5tO`*avFv $ %dXYj kF ouvfkdnn5tO`navFv  $ d>]nF $ 5Rc dF $U^opF ou ej '{N_ `  '>BBi==w w   s6I4*I99I>JJ#J J*J J)NN)FNN) r"r#r$r%model_input_namesr valid_kwargsrBILINEARr&dictstrrnfloatrCr@rUndarrayrr\rjBICUBICror rrSrrPILImager __classcell__)rFs@r*r/r/@sH"H((,L&*'9'B'B!#(&-!1504!W!W38nt#!W% !W  !W ! !W!W!We !W!WDK'$.!W4;&-!W !WN6:;? -zz-38n-++d2 - !11D8 -  -f %59;? (azz(a38n(a (a ++d2 (a !11D8 (a (a\(:'A'A59;? 0 zz0 38n0 % 0 ++d2 0 !11D8 0  0 l(:'A'A59;? %zz%38n%% % ++d2 % !11D8 % %N%&"&&*.2$(*."$"&'+$(150426/?/E/E;?#[C[C$;[C38nt# [C %t+ [C Tk [C!4K[Ct [C[C4K[C [CTk[CDK'$.[C4;&-[Cj(4/[C &,![C"!11D8#[C$ %[C'[Cr)r/)*r%numpyrUimage_processing_utilsrrrimage_transformsrrr r r image_utilsr r rrrrrrrrrrprocessing_utilsrutilsrrrutils.import_utilsrr get_loggerr"r}rrr/__all__r(r)r*rs'UU    -II?   H %  E  ;YC,YC YCx ! !r)