ΑiV+SrSSKrSSKrSSKrSSKrSSKr/rSr SSjr SSjr SSjr Sr SSjrSS jrSS jrSS jrSS jrSS jrg!\a SrNGf=f)at This file contains some common interfaces for image preprocess. Many users are confused about the image layout. We introduce the image layout as follows. - CHW Layout - The abbreviations: C=channel, H=Height, W=Width - The default layout of image opened by cv2 or PIL is HWC. PaddlePaddle only supports the CHW layout. And CHW is simply a transpose of HWC. It must transpose the input image. - Color format: RGB or BGR OpenCV use BGR color format. PIL use RGB color format. Both formats can be used for training. Noted that, the format should be keep consistent between the training and inference period. NcR[c SSKnURRS5 gg)NrzWarning with paddle image module: opencv-python should be imported, or paddle image module could NOT work; please install opencv-python first.FT)cv2sysstderrwrite)rs T/var/www/html/banglarbhumi/venv/lib/python3.13/site-packages/paddle/dataset/image.py _check_cv2r /s* {  W c ^US-nUSUS[R"53mUSUS[R"5S3n[RRT5(aU$[R"T5 [ R "U5nUR5n/n/n Sn UHn U RU;dMURURU 5R55 U RX+R5 [U5U:XdMqXS.n [R"U [ TSU 3S5S S 9 U S - n /n/n M [U5S:a)XS.n [R"U [ TSU 3S5S S 9 [ US S 9n U RU4Sj[R "T555 SSS5 U$!,(df  U$=f)a Read images from tar file and batch them into batch file. :param data_file: path of image tar file :type data_file: string :param dataset_name: 'train','test' or 'valid' :type dataset_name: string :param img2label: a dict with image file name as key and image's label as value :type img2label: dict :param num_per_batch: image number per batch file :type num_per_batch: int :return: path of list file containing paths of batch file :rtype: string _batch/_z.txtr)labeldataz/batch_wb)protocola)modec3r># UH,n[RRTSU35S-v M. g7f)r  N)ospathabspath).0fileout_paths r (batch_images_from_tar..qs5 , GGOOxj$0 1D 8,s47N)rgetpidrexistsmakedirstarfileopen getmembersnameappend extractfilereadlenpickledump writelineslistdir) data_file dataset_name img2label num_per_batch batch_dir meta_filetfmemsrlabelsfile_idmemoutputmetars @rbatch_images_from_tarr=<s$H$IAl^1RYY[M:H+Q|nAbiik]$?I ww~~h H i B ==?D D FG 88y KKs+002 3 MM)HH- .4yM)#)8 H:WWI6= 1  4y1}!0 DH:WWI6=  ic "d   8,   #  # " s 0G Gc[5SLdeU(aSOSn[R"[U5[RS9n[ R "X25nU$)a Load an color or gray image from bytes array. Example usage: .. code-block:: python >>> with open('cat.jpg') as f: ... im = load_image_bytes(f.read()) ... :param bytes: the input image bytes array. :type bytes: str :param is_color: If set is_color True, it will load and return a color image. Otherwise, it will load and return a gray image. :type is_color: bool Trrdtype)r npasarray bytearrayuint8rimdecode)bytesis_colorflag file_bytesimgs rload_image_bytesrKxsI$ <4  1aDIe,BHH=J ,,z (C Jr c[5SLdeU(aSOSn[R"URS5R S5U5nU$)aj Load an color or gray image from the file path. Example usage: .. code-block:: python >>> im = load_image('cat.jpg') :param file: the input image path. :type file: string :param is_color: If set is_color True, it will load and return a color image. Otherwise, it will load and return a gray image. :type is_color: bool Trrzutf-8)r rimreadencodedecode)rrGrHims r load_imagerQsH" <4  1aD DKK(//8$ ?B Ir c[5SLdeURSSup#XpTX#:aX-U-nOX-U-n[R"XU4[RS9nU$)a7 Resize an image so that the length of shorter edge is size. Example usage: .. code-block:: python >>> im = load_image('cat.jpg') >>> im = resize_short(im, 256) :param im: the input image with HWC layout. :type im: ndarray :param size: the shorter edge size of image after resizing. :type size: int TNr) interpolation)r shaperresize INTER_CUBIC)rPsizehwh_neww_news r resize_shortr\sb <4   88BQ>> im = load_image('cat.jpg') >>> im = resize_short(im, 256) >>> im = to_chw(im) :param im: the input image with HWC layout. :type im: ndarray :param order: the transposed order. :type order: tuple|list )r+rT transpose)rPorders rto_chwr`s1& rxx=CJ && & e B Ir cURSSup4X1- S-nXA- S-nXQ-Xa-pU(aXU2Xh2SS24nU$XU2Xh24nU$)aK Crop the center of image with size. Example usage: .. code-block:: python >>> im = load_image('cat.jpg') >>> im = center_crop(im, 224) :param im: the input image with HWC layout. :type im: ndarray :param size: the cropping size. :type size: int :param is_color: whether the image is color or not. :type is_color: bool Nr)rT rPrWrGrXrYh_startw_starth_endw_ends r center_croprgso$ 88BQ7>5  w}a/ 0 I w}, - Ir cURSSup4[RRSX1- S-5n[RRSXA- S-5nXQ-Xa-pU(aXU2Xh2SS24nU$XU2Xh24nU$)aL Randomly crop input image with size. Example usage: .. code-block:: python >>> im = load_image('cat.jpg') >>> im = random_crop(im, 224) :param im: the input image with HWC layout. :type im: ndarray :param size: the cropping size. :type size: int :param is_color: whether the image is color or not. :type is_color: bool Nrrr)rTrArandomrandintrbs r random_croprks$ 88BQ7>5  w}a/ 0 I w}, - Ir cx[UR5S:XaU(aUSS2SSS2SS24$USS2SSS24$)aV Flip an image along the horizontal direction. Return the flipped image. Example usage: .. code-block:: python >>> im = load_image('cat.jpg') >>> im = left_right_flip(im) :param im: input image with HWC layout or HW layout for gray image :type im: ndarray :param is_color: whether input image is color or not :type is_color: bool N)r+rT)rPrGs rleft_right_flipros<" 288}h!TrT1*~!TrT'{r cX[X5nU(a9[XUS9n[RR S5S:Xa [ X5nO [ XUS9n[UR5S:Xa [U5nURS5nUb[R"U[RS9nURS:Xa.U(a'USS2[R[R4nO7URS:XaUnO$[UR5[U5:XdeX-nU$) a Simply data argumentation for training. These operations include resizing, cropping and flipping. Example usage: .. code-block:: python >>> im = load_image('cat.jpg') >>> im = simple_transform(im, 256, 224, True) :param im: The input image with HWC layout. :type im: ndarray :param resize_size: The shorter edge length of the resized image. :type resize_size: int :param crop_size: The cropping size. :type crop_size: int :param is_train: Whether it is training or not. :type is_train: bool :param is_color: whether the image is color or not. :type is_color: bool :param mean: the mean values, which can be element-wise mean values or mean values per channel. :type mean: numpy array | list )rGrrrmfloat32Nr?r)r\rkrArirjrorgr+rTr`astypearrayrqndimnewaxis)rP resize_size crop_sizeis_trainrGmeans rsimple_transformrz4s8 b &B  : 99  Q 1 $ .B  : 288} BZ 9 B xxBJJ/ 99>h2::rzz12D YY!^Dtzz?c"g- --  Ir c6[X5n[XaX#XE5nU$)a Load image from the input file `filename` and transform image for data argumentation. Please refer to the `simple_transform` interface for the transform operations. Example usage: .. code-block:: python >>> im = load_and_transform('cat.jpg', 256, 224, True) :param filename: The file name of input image. :type filename: string :param resize_size: The shorter edge length of the resized image. :type resize_size: int :param crop_size: The cropping size. :type crop_size: int :param is_train: Whether it is training or not. :type is_train: bool :param is_color: whether the image is color or not. :type is_color: bool :param mean: the mean values, which can be element-wise mean values or mean values per channel. :type mean: numpy array | list )rQrz)filenamervrwrxrGryrPs rload_and_transformr}js!8 H 'B "9 OB Ir )i)T))rrr)TN)__doc__rr$numpyrAr ImportErrorr,__all__r r=rKrQr\r`rgrkrorzr}r rrs&   7;9x2<80::0?C3nEIE  CsAAA