kh#dZ ddlmcmZddlZdZdZejjj ddZ dZ dZ dZ d Zed fd Zed fd Zd d ed fdZGddZy)z Tensorflow Preprocessing Adapter Allows use of Tensorflow preprocessing pipeline in PyTorch Transform Copyright of original Tensorflow code below. Hacked together by / Copyright 2020 Ross Wightman N c tj|d||g5tjj|}tjj ||||||d}|\} } } tj | \} } } tj | \}}} tj | | ||g}tjj||d}|cdddS#1swYyxYw)aGenerates cropped_image using one of the bboxes randomly distorted. See `tf.image.sample_distorted_bounding_box` for more documentation. Args: image_bytes: `Tensor` of binary image data. bbox: `Tensor` of bounding boxes arranged `[1, num_boxes, coords]` where each coordinate is [0, 1) and the coordinates are arranged as `[ymin, xmin, ymax, xmax]`. If num_boxes is 0 then use the whole image. min_object_covered: An optional `float`. Defaults to `0.1`. The cropped area of the image must contain at least this fraction of any bounding box supplied. aspect_ratio_range: An optional list of `float`s. The cropped area of the image must have an aspect ratio = width / height within this range. area_range: An optional list of `float`s. The cropped area of the image must contain a fraction of the supplied image within in this range. max_attempts: An optional `int`. Number of attempts at generating a cropped region of the image of the specified constraints. After `max_attempts` failures, return the entire image. scope: Optional `str` for name scope. Returns: cropped image `Tensor` distorted_bounding_box_cropT)bounding_boxesmin_object_coveredaspect_ratio_range area_range max_attemptsuse_image_if_no_bounding_boxeschannelsN)tf name_scopeimageextract_jpeg_shapesample_distorted_bounding_boxunstackstackdecode_and_crop_jpeg) image_bytesbboxrr r r scopeshaper bbox_begin bbox_size_offset_yoffset_x target_height target_width crop_windowrs V/var/www/teggl/fontify/venv/lib/python3.12/site-packages/timm/data/tf_preprocessing.pyrr!s> u;k4=P Q++K8(*(N(N 11!%+/)O)1%$A Iq!# : 6(A)+I)>& |Qhh(M<PQ --k;QR-S%s B:CC'ctj||}tj|tj}tjtj ||S)z0At least `x` of `a` and `b` `Tensors` are equal.)requalcastint32 greater_equal reduce_sum)abxmatchs r$_at_least_x_are_equalr/UsA HHQNE GGE288 $E  BMM%0! 44c Dtjgdtjgd}t|ddddd tjj }t |tjd }tj|fd fd S) z!Make a random crop of image_size.)r2?r3)r4)dtyper皙?)?gUUUUUU?)g{Gz?r3 N)rr r r rr ctS)N)_decode_and_center_crop)r image_sizesr$z)_decode_and_random_crop..ls' Z@r0cRtjjggdSNr)rrresize)rr< resize_methodsr$r=z)_decode_and_random_crop..ms%*j)A=QRSTr0) rconstantfloat32rrrr/rcond)rr<rAroriginal_shapebadrs``` @r$_decode_and_random_croprG\s ;;+2::Y OD ' , EXX00=N  CC GG @T VE Lr0c tjj|}|d}|d}tj||tzz tjtj ||tj ztj}||z dzdz}||z dzdz}tj||||g} tjj|| d} tjj| g||g|d} | S)z=Crops to center of image with padding then scales image_size.rr4r r) rrrr' CROP_PADDINGminimumrCr(rrr@) rr<rAr image_height image_widthpadded_center_crop_size offset_height offset_widthr#rs r$r;r;rs HH ' ' 4E8L(K gg  \1 2 L+6 C D  #%<<AaGM #::a?AEL((M<35LNOK HH ) )+{Q ) OE HHOOUGj*%=} Ma PE Lr0cDtjj|}|S)zRandom horizontal image flip.)rrrandom_flip_left_right)rs r$_fliprSs HH + +E 2E Lr0bicubicc|dk(r$tjjjn#tjjj}t |||}t |}tj|||dg}tjj||rtjntj}|SaAPreprocesses the given image for evaluation. Args: image_bytes: `Tensor` representing an image binary of arbitrary size. use_bfloat16: `bool` for whether to use bfloat16. image_size: image size. interpolation: image interpolation method Returns: A preprocessed image `Tensor`. rTr )r6) rr ResizeMethodBICUBICBILINEARrGrSreshapeconvert_image_dtypebfloat16rCr use_bfloat16r< interpolationrArs r$preprocess_for_trainr`s6Ci5OBHH))11UWU]U]UjUjUsUsM #K] KE %LE JJuz:q9 :E HH ( ( LR[[bjj ) BE Lr0cn|dk(r$tjjjn#tjjj}t |||}tj |||dg}tjj||rtjntj}|SrV) rrrWrXrYr;rZr[r\rCr]s r$preprocess_for_evalrbs6Ci5OBHH))11UWU]U]UjUjUsUsM #K] KE JJuz:q9 :E HH ( ( LR[[bjj ) BE Lr0Fc>|rt||||St||||S)aPreprocesses the given image. Args: image_bytes: `Tensor` representing an image binary of arbitrary size. is_training: `bool` for whether the preprocessing is for training. use_bfloat16: `bool` for whether to use bfloat16. image_size: image size. interpolation: image interpolation method Returns: A preprocessed image `Tensor` with value range of [0, 255]. )r`rb)r is_trainingr^r<r_s r$preprocess_imageres)"#Kz=YY"; j-XXr0c eZdZddZdZdZy)TfPreprocessTransformc||_t|tr|dn||_||_d|_|j |_d|_yr?) rd isinstancetuplesizer_ _image_bytes_build_tf_graph process_imagesess)selfrdrkr_s r$__init__zTfPreprocessTransform.__init__sH&)$6DGD * !113 r0ctjd5tjgtj|_t |j|j d|j|j}ddd|S#1swYSxYw)Nz/cpu:0)rr6F) rdevice placeholderstringrlrerdrkr_)rpimgs r$rmz%TfPreprocessTransform._build_tf_graphsx YYx  [ "ii!D #!!4#3#3UDIItGYGY[C  [  [ s A"BB c|jtj|_|jj|j|j |i}|j jddjtj}|jdkrtj|d}tj|d}|S)N) feed_dictrr )axisrI)rorSessionrunrnrlroundclipastypenpuint8ndim expand_dimsrollaxis)rprrvs r$__call__zTfPreprocessTransform.__call__s 99  DIiimmD..4;L;Lk:Zm[iikq#&--bhh7 88a<..2.Ckk#q! r0N)FrrT)__name__ __module__ __qualname__rqrmrr0r$rgrgsr0rg)r7)r8gHzG?)g?r3dN)__doc__tensorflow.compat.v1compatv1rnumpyr IMAGE_SIZErJdisable_eager_executionrr/rGr;rSr`rbrergrr0r$rs.*!!    $$&473?+6-0&* 1h5,* @JYb*?IXa*"'"' *#, Y.r0