o hj@s6ddlZddlmZmZmZmZmZddlZddl m Z m Z m Z m Z mZddlmZmZmZmZmZddlmZmZmZmZmZerMddlZddl mZerTddlZer[ddlZerdddl mZ! dRdej"d ee e#fd eee e#fd ej"fd d Z$dej%dfdej"de&dee dej'deee#e fd ej"f ddZ(ddZ)  dSdeej"ddddfdee*deee#e fd dfddZ+   dTdej"dee,ee,e,fee,ee,fd e*d!ee,deee#e fd e-f d"d#Z.     dUdej"dee,e,fd$d%d&ee,dee d'e*deee#e fd ej"fd(d)Z/  dSdej"d*ee&ee&fd+ee&ee&fdee deee#e fd ej"f d,d-Z0   dVdej"dee,e,fdeee#e fdeee#e fd'ee*d ej"f d.d/Z1dWd1d2Z2d0ej"d ej"fd3d4Z3dXd5d6Z4d0ed efd7d8Z5dYd:d;Z6d9ej"d ej"fdd?Z8d9ed efd@dAZ9dBdCZ:dDdEZ;GdFdGdGeZde d e fdNdOZ?  dSdej"dee deee#e fd ej"fdPdQZ@dS)[N)IterableListOptionalTupleUnion)ChannelDimension ImageInputget_channel_dimension_axisget_image_sizeinfer_channel_dimension_format) ExplicitEnum TensorType is_jax_tensor is_tf_tensoris_torch_tensor)is_flax_availableis_tf_availableis_torch_availableis_vision_availablerequires_backends)PILImageResamplingimage channel_diminput_channel_dimreturncCst|tjstdt||durt|}t|}||kr!|S|tjkr-|d}|S|tj kr9|d}|Std |)a) Converts `image` to the channel dimension format specified by `channel_dim`. Args: image (`numpy.ndarray`): The image to have its channel dimension set. channel_dim (`ChannelDimension`): The channel dimension format to use. input_channel_dim (`ChannelDimension`, *optional*): The channel dimension format of the input image. If not provided, it will be inferred from the input image. Returns: `np.ndarray`: The image with the channel dimension set to `channel_dim`. ,Input image must be of type np.ndarray, got N)rr)rrrz(Unsupported channel dimension format: {}) isinstancenpndarray ValueErrortyper rFIRST transposeLASTformat)rrrtarget_channel_dimr(S/var/www/html/ai/venv/lib/python3.10/site-packages/transformers/image_transforms.pyto_channel_dimension_format5s    r*scale data_formatdtypeinput_data_formatcCsHt|tjstdt|||}|durt|||}||}|S)a Rescales `image` by `scale`. Args: image (`np.ndarray`): The image to rescale. scale (`float`): The scale to use for rescaling the image. data_format (`ChannelDimension`, *optional*): The channel dimension format of the image. If not provided, it will be the same as the input image. dtype (`np.dtype`, *optional*, defaults to `np.float32`): The dtype of the output image. Defaults to `np.float32`. Used for backwards compatibility with feature extractors. input_data_format (`ChannelDimension`, *optional*): The channel dimension format of the input image. If not provided, it will be inferred from the input image. Returns: `np.ndarray`: The rescaled image. rN)rrr r!r"r*astype)rr+r,r-r.rescaled_imager(r(r)rescale\s   r1cCs|jtjkr d}|St||tr4td|kr%t|dkr%d}|Std|d| dtd|krFt|dkrFd}|Std |d| d) z Detects whether or not the image needs to be rescaled before being converted to a PIL image. The assumption is that if the image is of type `np.float` and all values are between 0 and 1, it needs to be rescaled. FrzZThe image to be converted to a PIL image contains values outside the range [0, 255], got [z, z%] which cannot be converted to uint8.rTzXThe image to be converted to a PIL image contains values outside the range [0, 1], got [) r-ruint8allcloser/intallr!minmax)r do_rescaler(r(r)_rescale_for_pil_conversions0  r:zPIL.Image.Image torch.Tensor tf.Tensorz jnp.ndarrayr9cCsttdgt|tjjr|St|st|r|}nt|r&t |}nt|t j s5t d t|t|tj|}|jddkrJt j|ddn|}|durTt|n|}|r]t|d}|t j}tj|S)aI Converts `image` to a PIL Image. Optionally rescales it and puts the channel dimension back as the last axis if needed. Args: image (`PIL.Image.Image` or `numpy.ndarray` or `torch.Tensor` or `tf.Tensor`): The image to convert to the `PIL.Image` format. do_rescale (`bool`, *optional*): Whether or not to apply the scaling factor (to make pixel values integers between 0 and 255). Will default to `True` if the image type is a floating type and casting to `int` would result in a loss of precision, and `False` otherwise. input_data_format (`ChannelDimension`, *optional*): The channel dimension format of the input image. If unset, will use the inferred format from the input. Returns: `PIL.Image.Image`: The converted image. visionz"Input image type not supported: {}raxisNr2)r to_pil_imagerPILImagerrnumpyrrarrayr r!r&r"r*rr%shapesqueezer:r1r/r3 fromarray)rr9r.r(r(r)rAs        rAT input_imagesizedefault_to_squaremax_sizec Cst|ttfr t|dkrt|St|dkr|d}ntd|r&||fSt||\}}||kr5||fn||f\}}|} | t| ||} } |duri|| krZtd|d|| |krit|| | |} } ||krq| | fS| | fS)a Find the target (height, width) dimension of the output image after resizing given the input image and the desired size. Args: input_image (`np.ndarray`): The image to resize. size (`int` or `Tuple[int, int]` or List[int] or Tuple[int]): The size to use for resizing the image. If `size` is a sequence like (h, w), output size will be matched to this. If `size` is an int and `default_to_square` is `True`, then image will be resized to (size, size). If `size` is an int and `default_to_square` is `False`, then smaller edge of the image will be matched to this number. i.e, if height > width, then image will be rescaled to (size * height / width, size). default_to_square (`bool`, *optional*, defaults to `True`): How to convert `size` when it is a single int. If set to `True`, the `size` will be converted to a square (`size`,`size`). If set to `False`, will replicate [`torchvision.transforms.Resize`](https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Resize) with support for resizing only the smallest edge and providing an optional `max_size`. max_size (`int`, *optional*): The maximum allowed for the longer edge of the resized image: if the longer edge of the image is greater than `max_size` after being resized according to `size`, then the image is resized again so that the longer edge is equal to `max_size`. As a result, `size` might be overruled, i.e the smaller edge may be shorter than `size`. Only used if `default_to_square` is `False`. input_data_format (`ChannelDimension`, *optional*): The channel dimension format of the input image. If unset, will use the inferred format from the input. Returns: `tuple`: The target (height, width) dimension of the output image after resizing. rrrz7size must have 1 or 2 elements if it is a list or tupleNz max_size = zN must be strictly greater than the requested size for the smaller edge size = )rtuplelistlenr!r r5) rIrJrKrLr.heightwidthshortlongrequested_new_short new_shortnew_longr(r(r)get_resize_output_image_sizes,%   rWresampler reducing_gap return_numpyc Csttdg|dur |ntj}t|dkstd|dur!t|}|dur'|n|}d}t|tj j s=t |}t |||d}|\}} |j| |f||d} |rqt | } | jdkr^t j| dd n| } t| |tjd } |rot| d n| } | S) a Resizes `image` to `(height, width)` specified by `size` using the PIL library. Args: image (`np.ndarray`): The image to resize. size (`Tuple[int, int]`): The size to use for resizing the image. resample (`int`, *optional*, defaults to `PILImageResampling.BILINEAR`): The filter to user for resampling. reducing_gap (`int`, *optional*): Apply optimization by resizing the image in two steps. The bigger `reducing_gap`, the closer the result to the fair resampling. See corresponding Pillow documentation for more details. data_format (`ChannelDimension`, *optional*): The channel dimension format of the output image. If unset, will use the inferred format from the input. return_numpy (`bool`, *optional*, defaults to `True`): Whether or not to return the resized image as a numpy array. If False a `PIL.Image.Image` object is returned. input_data_format (`ChannelDimension`, *optional*): The channel dimension format of the input image. If unset, will use the inferred format from the input. Returns: `np.ndarray`: The resized image. r=Nrzsize must have 2 elementsF)r9r.)rXrYr>r?rgp?)rresizerBILINEARrOr!r rrBrCr:rArrEndim expand_dimsr*rr%r1) rrJrXrYr,rZr.r9rPrQ resized_imager(r(r)r\s* !  r\meanstdcCs2t|tjs td|durt|}t||d}|j|}t|jtj s+| tj }t|t rCt ||krBtd|dt |n|g|}tj||jd}t|t rht ||krgtd|dt |n|g|}tj||jd}|tjkr|||}n|j||j}|durt|||}|S|}|S)a Normalizes `image` using the mean and standard deviation specified by `mean` and `std`. image = (image - mean) / std Args: image (`np.ndarray`): The image to normalize. mean (`float` or `Iterable[float]`): The mean to use for normalization. std (`float` or `Iterable[float]`): The standard deviation to use for normalization. data_format (`ChannelDimension`, *optional*): The channel dimension format of the output image. If unset, will use the inferred format from the input. input_data_format (`ChannelDimension`, *optional*): The channel dimension format of the input image. If unset, will use the inferred format from the input. zimage must be a numpy arrayN)r.zmean must have z$ elements if it is an iterable, got r-zstd must have )rrr r!r r rF issubdtyper-floatingr/float32rrOrErr%Tr*)rrarbr,r. channel_axis num_channelsr(r(r) normalize[s6           rjcCsttdg|durtdt|durdn|}t|tjs'tdt |t|t r2t |dkr6td|dur>t |}|durD|n|}t |tj|}t|tj\}}|\}} t|t| }} ||d} | |} || d} | | } | dkr| |kr| dkr| |kr|d | | | | f}t ||tj}|St||}t| |}|jdd ||f}tj||d }||d}||}||d}||}||d ||||f<| |7} | |7} | |7} | |7} |d td| t|| td| t|| f}t ||tj}|s t|}|S) a Crops the `image` to the specified `size` using a center crop. Note that if the image is too small to be cropped to the size given, it will be padded (so the returned result will always be of size `size`). Args: image (`np.ndarray`): The image to crop. size (`Tuple[int, int]`): The target size for the cropped image. data_format (`str` or `ChannelDimension`, *optional*): The channel dimension format for the output image. Can be one of: - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format. If unset, will use the inferred format of the input image. input_data_format (`str` or `ChannelDimension`, *optional*): The channel dimension format for the input image. Can be one of: - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format. If unset, will use the inferred format of the input image. return_numpy (`bool`, *optional*): Whether or not to return the cropped image as a numpy array. Used for backwards compatibility with the previous ImageFeatureExtractionMixin method. - Unset: will return the same type as the input image. - `True`: will return a numpy array. - `False`: will return a `PIL.Image.Image` object. Returns: `np.ndarray`: The cropped image. r=Nz8return_numpy is deprecated and will be removed in v.4.33TrrzOsize must have 2 elements representing the height and width of the output imager.)rF)r center_cropwarningswarn FutureWarningrrr r!r"rrOr r*rr#r r5r8rF zeros_liker7rA)rrJr,r.rZoutput_data_format orig_height orig_width crop_height crop_widthtopbottomleftright new_height new_width new_shape new_imagetop_pad bottom_padleft_pad right_padr(r(r)rlsR #         .rl bboxes_centercCsL|d\}}}}tj|d||d||d||d|gdd}|S)Nr>?dimunbindtorchstack)rcenter_xcenter_yrQrP bbox_cornersr(r(r)_center_to_corners_format_torchs *rcCsH|j\}}}}tj|d||d||d||d|gdd}|S)Nrr>r?rgrrrrrrQrPbboxes_cornersr(r(r)_center_to_corners_format_numpys *rcCsPtj|dd\}}}}tj|d||d||d||d|gdd}|S)Nr>r?rtfunstackrrr(r(r)_center_to_corners_format_tfs *rcCFt|rt|St|tjrt|St|rt|Stdt |)a| Converts bounding boxes from center format to corners format. center format: contains the coordinate for the center of the box and its width, height dimensions (center_x, center_y, width, height) corners format: contains the coodinates for the top-left and bottom-right corners of the box (top_left_x, top_left_y, bottom_right_x, bottom_right_y) Unsupported input type ) rrrrr rrrr!r")rr(r(r)center_to_corners_formats  rrcCsD|d\}}}}||d||d||||g}tj|ddS)Nr>rrr)r top_left_x top_left_ybottom_right_xbottom_right_ybr(r(r)_corners_to_center_format_torch.s  rcCs@|j\}}}}tj||d||d||||gdd}|S)Nrr>r?rrrrrrrr(r(r)_corners_to_center_format_numpy9s   rcCsHtj|dd\}}}}tj||d||d||||gdd}|S)Nr>r?rrrr(r(r)_corners_to_center_format_tfGs   rcCr)a Converts bounding boxes from corners format to center format. corners format: contains the coordinates for the top-left and bottom-right corners of the box (top_left_x, top_left_y, bottom_right_x, bottom_right_y) center format: contains the coordinate for the center of the box and its the width, height dimensions (center_x, center_y, width, height) r) rrrrr rrrr!r")rr(r(r)corners_to_center_formatUs  rcCst|tjr>t|jdkr>|jtjkr|tj}|dddddfd|dddddfd|dddddfSt |dd|dd|dS)z* Converts RGB color to unique ID. Nrrir) rrr rOrFr-r3r/int32r5)colorr(r(r) rgb_to_idls   J$rcCst|tjr1|}tt|jdg}tj|tjd}t dD]}|d|d|f<|d}q |Sg}t dD] }| |d|d}q7|S)z* Converts unique ID to RGB color. rrcr.) rrr copyrMrNrFzerosr3rangeappend)id_map id_map_copy rgb_shapergb_mapir_r(r(r) id_to_rgbws     rc@s eZdZdZdZdZdZdZdS) PaddingModezP Enum class for the different padding modes to use when padding images. constantreflect replicate symmetricN)__name__ __module__ __qualname____doc__CONSTANTREFLECT REPLICATE SYMMETRICr(r(r(r)rs rgpaddingmodeconstant_valuescsdurtfdd}||}|tjkr&||}tj|d|dn1|tjkr4tj|ddn#|tjkrBtj|ddn|tjkrPtj|d dntd ||durct |SS) a Pads the `image` with the specified (height, width) `padding` and `mode`. Args: image (`np.ndarray`): The image to pad. padding (`int` or `Tuple[int, int]` or `Iterable[Tuple[int, int]]`): Padding to apply to the edges of the height, width axes. Can be one of three formats: - `((before_height, after_height), (before_width, after_width))` unique pad widths for each axis. - `((before, after),)` yields same before and after pad for height and width. - `(pad,)` or int is a shortcut for before = after = pad width for all axes. mode (`PaddingMode`): The padding mode to use. Can be one of: - `"constant"`: pads with a constant value. - `"reflect"`: pads with the reflection of the vector mirrored on the first and last values of the vector along each axis. - `"replicate"`: pads with the replication of the last value on the edge of the array along each axis. - `"symmetric"`: pads with the reflection of the vector mirrored along the edge of the array. constant_values (`float` or `Iterable[float]`, *optional*): The value to use for the padding if `mode` is `"constant"`. data_format (`str` or `ChannelDimension`, *optional*): The channel dimension format for the output image. Can be one of: - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format. If unset, will use same as the input image. input_data_format (`str` or `ChannelDimension`, *optional*): The channel dimension format for the input image. Can be one of: - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format. If unset, will use the inferred format of the input image. Returns: `np.ndarray`: The padded image. Ncst|ttfr||f||ff}nOt|tr,t|dkr,|d|df|d|dff}n3t|trCt|dkrCt|dtrC||f}nt|trXt|dkrXt|dtrX|}ntd|tjkrjdg|Rng|dR}jdkr~dg|R}|S|}|S)za Convert values to be in the format expected by np.pad based on the data format. rrrzUnsupported format: )rr) rr5floatrMrOr!rr#r^)valuesrr.r(r)_expand_for_data_formats"$ $$z$pad.._expand_for_data_formatr)rrr)redgerzInvalid padding mode: ) r rrrpadrrrr!r*)rrrrr,r.rr(rr)rs$+    rcCs,ttdgt|tjjs|S|d}|S)z Converts an image to RGB format. Only converts if the image is of type PIL.Image.Image, otherwise returns the image as is. Args: image (Image): The image to convert. r=RGB)rconvert_to_rgbrrBrCconvert)rr(r(r)rs  rcCsx|durt|n|}|tjkr|ddddf}n|tjkr(|ddddf}ntd||dur:t|||d}|S)a Flips the channel order of the image. If the image is in RGB format, it will be converted to BGR and vice versa. Args: image (`np.ndarray`): The image to flip. data_format (`ChannelDimension`, *optional*): The channel dimension format for the output image. Can be one of: - `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `ChannelDimension.LAST`: image in (height, width, num_channels) format. If unset, will use same as the input image. input_data_format (`ChannelDimension`, *optional*): The channel dimension format for the input image. Can be one of: - `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `ChannelDimension.LAST`: image in (height, width, num_channels) format. If unset, will use the inferred format of the input image. N.r>zUnsupported channel dimension: r[)r rr%r#r!r*)rr,r.r(r(r)flip_channel_orders  r)N)NN)TNN)NNNTN)NNN)rr;rr;)rr<rr<)rr;rr;)rr<rr<)ArmtypingrrrrrrDr image_utilsrr r r r utilsr rrrrutils.import_utilsrrrrrrBrr tensorflowr jax.numpyjnpr strr*rfrr-r1r:boolrAr5rMrWr\rjrlrrrrrrrrrrrrrrrr(r(r(r)sH     * & 7 F  K ?  c      W