^j(ddlZddlZddlmZddlmZddlmZm Z m Z ddl m Z ddl mZmZmZmZddlmZmZdd lmZmZGd d ed ZeGddeZdgZy)N) functional)TorchvisionBackend) BatchFeatureget_patch_output_sizeselect_best_resolution)divide_to_patches)ChannelDimensionPILImageResamplingSizeDictget_image_size) ImagesKwargsUnpack) TensorTypeauto_docstringcFeZdZUdZeed<eed<eeeed<eed<y)AriaImageProcessorKwargsa max_image_size (`int`, *optional*, defaults to `self.max_image_size`): Maximum image size. Must be either 490 or 980. min_image_size (`int`, *optional*, defaults to `self.min_image_size`): Minimum image size. Images smaller than this in any dimension will be scaled up. split_resolutions (`list[list[int]]`, *optional*, defaults to `self.split_resolutions`): A list of possible resolutions as (height, width) pairs for splitting high-resolution images into patches. split_image (`bool`, *optional*, defaults to `self.split_image`): Whether to split the image into patches using the best matching resolution from `split_resolutions`. max_image_sizemin_image_sizesplit_resolutions split_imageN)__name__ __module__ __qualname____doc__int__annotations__listbooly/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/aria/image_processing_aria.pyrr!s) DI&r!rF)totalceZdZgdZeZejZgdZ gdZ dZ dZ dZ dZdZdZdZdeeffd Zd ed ed eefd Zddd eddd dfdZddd ed dfdZdddeeededdd edf dZ d'deddedededeeezdzdeeezdzdedzdeezdzded ed!eeedzd"eddd efd#Z d(d$ed%efd&Z!xZ"S))AriaImageProcessor pixel_values pixel_mask num_crops)?r*r*PFNTkwargsc |jd%gd}|Dcgc]}|ddz|ddzgc}|d<t|di|ycc}w)Nr)))r/r)r/)r/)r/)r/)r/)r0r1)r0r)r0r0)r0r/)rr/)rr0)r1r/)r1r0)r2r/)r3r/)r4r/)r5r/rr/r )getsuper__init__)selfr-default_resolutionsel __class__s r"r9zAriaImageProcessor.__init__Cs] ::) * 2#{ Pc*d"BqECKA+E*dF& ' "6"+esAoriginal_resolutiontarget_resolutionreturncv|\}}|\}}t||z d\}}t||z d\} } || ||z| | zgS)zNGet padding size for patching, returns [left, top, right, bottom] for tvF.pad.r0)divmod) r:r>r?original_heightoriginal_width target_height target_widthpaste_xr_xpaste_yr_ys r"_get_padding_sizez$AriaImageProcessor._get_padding_sizeIsW*='&7# |l^;Q? mo=qA 'C-3??r!imagez torch.Tensorresamplez7PILImageResampling | tvF.InterpolationMode | int | Nonec|t||tj\}}|j|t |||S)zFResize an image to a target resolution while maintaining aspect ratio.input_data_formatheightwidth)rr FIRSTresizer )r:rLr?rM new_height new_widths r"_resize_for_patchingz'AriaImageProcessor._resize_for_patchingQs?!6 $8H8N8N!  I{{5(*I"NPXYYr!ct||tj}|j||}t j ||S)zCPad an image to a target resolution while maintaining aspect ratio.rO)padding)rr rTrKtvFpad)r:rLr?new_resolutionrZs r"_pad_for_patchingz$AriaImageProcessor._pad_for_patching]s> /u6G[k[q[qr((9JKwwug..r!grid_pinpoints patch_sizect|ts tdt|tj }t ||}|j|||}|j||}t||} | S)a Process an image with variable resolutions by dividing it into patches. Args: image (`torch.Tensor`): The input image to be processed (channels-first format). grid_pinpoints (`list[list[int]]`): A list of possible resolutions as (height, width) pairs. patch_size (`int`): Size of each square patch to divide the image into. resample (`PILImageResampling | tvF.InterpolationMode | int | None`): Resampling filter to use when resizing. Returns: `list[torch.Tensor]`: A list of image patches in channels-first format. z6grid_pinpoints must be a list of possible resolutions.) channel_dim)r`) isinstancer TypeErrorr r rTrrXr^r ) r:rLr_r`rM image_sizebest_resolution resized_image padded_imagepatchess r"get_image_patchesz$AriaImageProcessor.get_image_patchesgsp..$/TU U#E7G7M7MN 0^L11%(S --m_M #LZHr!images do_rescalerescale_factor do_normalize image_mean image_stddisable_groupingreturn_tensorsrrrrc ^| dvr tdg}g}d}|D]>}| r|j|| | | }n|g}|t||kDr t|}|D]}|jd|jd}}| t ||z }||k\rt t ||z| }| }n| }t t ||z| }|j |t||| }| |z }| |z }tj|dd||g}tj| | ftj}d|d|d|f<|j||j|Atj|d }|j||||||}tj|d }t!|||d | S) N)r6r+z(max_image_size must be either 490 or 980rQr)dtypeT)dimr&)data tensor_type) ValueErrorrjlenshapemaxrrUr r[r\torchzerosrappendstackrescale_and_normalizer)r:rkrlrmrnrorprqrrrrrrrMr- pixel_masksprocessed_cropsr)rL crop_images crop_imagehwscalenew_hnew_wpadding_bottom padding_rightr(stacked_images stacked_maskss r" _preprocesszAriaImageProcessor._preprocesss"  +GH H   3E"44U*"":.&&z2' 3 3<_!<33 J jR[  KQ7  .+& '   r!rRrSc>|jd|j}|jd|j}|jd|j}t ||f|\}}|sd} | St j ||z t j ||z z} | S)a A utility that returns number of image patches for a given image size. Args: height (`int`): Height of the input image. width (`int`): Width of the input image. images_kwargs (`dict`, *optional*): Any kwargs to override defaults of the image processor. Returns: `int`: Number of patches per image. rrrr/)r7rrrrmathceil) r:rRrS images_kwargsrrrresized_height resized_width num_patchess r"get_number_of_image_patchesz.AriaImageProcessor.get_number_of_image_patchess$'' t7G7GH &**+;T=P=PQ)--.A4CYCYZ(>Pa(b%    >N:;dii XfHf>gg  r!)r+r,NFN)N)#rrrmodel_input_namesr valid_kwargsr BICUBICrMrorprrrrdo_convert_rgbrlrnrr9tuplerrrKrXr^rjrfloatstrrrrr __classcell__)r=s@r"r%r%3sC+L!))H JINNKNJL#(@!A# @U@u@Y]^aYb@ Z Z! ZL Z  Z//!/  /T#Y  L  n  V"!48!NRC ^$C C  C  C DK'$. C 4;&-C +C j(4/C C C  S ?T1C C LC !C J#cr!r%)rr~torchvision.transforms.v2rr[image_processing_backendsrimage_processing_utilsrrrimage_transformsr image_utilsr r r r processing_utilsrrutilsrrrr%__all__r r!r"rs`( 7;aa1YY4/|5$r+rrj r!