qyiFSSKrSSKJr SSKrSSKJs Js Jr SSK J r SSK J r SSK JrJrJr SSKJrJrJrJr SSKJrJr SS KJrJr \(aS S KJr S rS rSr Sr!\"SS\ 55r"S/r#g)N) TYPE_CHECKING)TorchvisionBackend) BatchFeature)center_to_corners_formatgroup_images_by_shapereorder_images)OPENAI_CLIP_MEANOPENAI_CLIP_STDPILImageResamplingSizeDict) ImagesKwargsUnpack) TensorTypeauto_docstring)Owlv2ObjectDetectionOutputc,UR5(a@UR[R[R4;aU$UR 5$UR[R [R4;aU$UR5$)N) is_floating_pointdtypetorchfloat32float64floatint32int64int)ts m/home/mande/repo/quber/.venv/lib/python3.13/site-packages/transformers/models/owlv2/image_processing_owlv2.py_upcastr 'scGG u}}==qL1779LGG U[[99qFquuwFcf[U5nUSS2S4USS2S4- USS2S4USS2S4- -$)a Computes the area of a set of bounding boxes, which are specified by its (x1, y1, x2, y2) coordinates. Args: boxes (`torch.FloatTensor` of shape `(number_of_boxes, 4)`): Boxes for which the area will be computed. They are expected to be in (x1, y1, x2, y2) format with `0 <= x1 < x2` and `0 <= y1 < y2`. Returns: `torch.FloatTensor`: a tensor containing the area for each box. Nrrr)r )boxess rbox_arear%/sB ENE !Q$K%1+ %%1+ad *C DDr!cV[U5n[U5n[R"USS2SSS24USS2SS245n[R"USS2SSS24USS2SS245nXT- R SS9nUSS2SS2S4USS2SS2S4-nUSS2S4U-U- nXx- n X4$)Nr#rminr)r%rmaxr(clamp) boxes1boxes2area1area2left_top right_bottom width_heightinterunionious rbox_iour5>s V E V Eyy4! ,fQUm[U[R 5(aUR S5up4O [S5e[R"X45n[R"XUXU/SS9nURS5RUR5nX-nU$s snfs snf)a Scale batch of bounding boxes to the target sizes. Args: boxes (`torch.Tensor` of shape `(batch_size, num_boxes, 4)`): Bounding boxes to scale. Each box is expected to be in (x1, y1, x2, y2) format. target_sizes (`list[tuple[int, int]]` or `torch.Tensor` of shape `(batch_size, 2)`): Target sizes to scale the boxes to. Each target size is expected to be in (height, width) format. Returns: `torch.Tensor` of shape `(batch_size, num_boxes, 4)`: Scaled bounding boxes. rrz4`target_sizes` must be a list, tuple or torch.Tensordim) isinstancelisttuplertensorTensorunbind TypeErrorr)stack unsqueezetodevice)r$ target_sizesi image_height image_widthmax_size scale_factors r _scale_boxesrJNs,u ..||<$@SU;aURS5nX!S'[TU]"S0UD6 g)Nrescale do_rescale)popsuper__init__)selfrQ rescale_val __class__s rrXOwlv2ImageProcessor.__init__~s3   **Y/K#.<  "6"r!outputsr thresholdrDcURURpT[U5nUb[U5U:wa [S5e[R "USS9n[R "UR5nURn [U5nUb [XS5n/n [XU5H*upn X:nXn Xn Xn U RXU S.5 M, U $)a Converts the raw output of [`Owlv2ForObjectDetection`] into final bounding boxes in (top_left_x, top_left_y, bottom_right_x, bottom_right_y) format. Args: outputs ([`Owlv2ObjectDetectionOutput`]): Raw outputs of the model. threshold (`float`, *optional*, defaults to 0.1): Score threshold to keep object detection predictions. target_sizes (`torch.Tensor` or `list[tuple[int, int]]`, *optional*): Tensor of shape `(batch_size, 2)` or list of tuples (`tuple[int, int]`) containing the target size `(height, width)` of each image in the batch. If unset, predictions will not be resized. Returns: `list[Dict]`: A list of dictionaries, each dictionary containing the following keys: - "scores": The confidence scores for each predicted box on the image. - "labels": Indexes of the classes predicted by the model on the image. - "boxes": Image bounding boxes in (top_left_x, top_left_y, bottom_right_x, bottom_right_y) format. z9Make sure that you pass in as many target sizes as imagesr7scoreslabelsr$) logits pred_boxeslen ValueErrorrr)sigmoidvaluesindicesrrJzipappend)rYr]r^rD batch_logits batch_boxes batch_sizebatch_class_logits batch_scores batch_labelsresultsrbrcr$keeps rpost_process_object_detection1Owlv2ImageProcessor.post_process_object_detections2%,NNG4F4Fk&  #L(9Z(GXY Y#YY|<}}%7%>%>? )11 /{;   #&{AK%([%Q !FE%D\F\FKE NNfO P &Rr!cURURpeUb#[U5[U5:wa [S5eUbURSS:wa [S5e[ R "USS9n[ R"UR5n[U5nUS:a[URS 5Hjn [ R"X*5HJn XU (dM[XiU SS24RS 5Xi5S S n S X'S XX:'ML Ml Ub [Xd5n/n [ R"U5n [URS 5Hn XnUR!5R#5(dM,S XU:'[ R "U5S -nXS -- US-- n[ R$"US S5nUX'XS :nXUnXiUnU R'USUS.5 M U $)a2 Converts the output of [`Owlv2ForObjectDetection.image_guided_detection`] into the format expected by the COCO api. Args: outputs ([`Owlv2ImageGuidedObjectDetectionOutput`]): Raw outputs of the model. threshold (`float`, *optional*, defaults to 0.0): Minimum confidence threshold to use to filter out predicted boxes. nms_threshold (`float`, *optional*, defaults to 0.3): IoU threshold for non-maximum suppression of overlapping boxes. target_sizes (`torch.Tensor`, *optional*): Tensor of shape (batch_size, 2) where each entry is the (height, width) of the corresponding image in the batch. If set, predicted normalized bounding boxes are rescaled to the target sizes. If left to None, predictions will not be unnormalized. Returns: `list[Dict]`: A list of dictionaries, each dictionary containing the scores, labels and boxes for an image in the batch as predicted by the model. All labels are set to None as `Owlv2ForObjectDetection.image_guided_detection` perform one-shot object detection. NzTMake sure that you pass in as many target sizes as the batch dimension of the logitsrr#zTEach element of target_sizes must contain the size (h, w) of each image of the batchr`r7g?rggư>皙?g?ra)rdtarget_pred_boxesrfrgshaperr)rhrirrangeargsortr5rArJ zeros_likenonzeronumelcliprl)rYr]r^ nms_thresholdrDrd target_boxesprobsrbidxrEiousrsalphas query_scores max_score query_alphasmask box_scoresr$s r#post_process_image_guided_detection7Owlv2ImageProcessor.post_process_image_guided_detections , '~~w/H/H  #F s<7H(Hst t  # (:(:1(=(Bst t &b)u||,0 =  3 \//23 |4A!;q> "<#4QT#:#D#DQ#GIZ[\]^_`aD"DG8;FK 45 54  #' CL!!&)++A./C!;L'')//1169L 1 2 ,/$6I(O<SQL ::lC=L&FK;?DT*J %d+E NNjD5Q R'0*r!imagesz torch.Tensorconstant_valuereturncURSSup4[X45nXS- nXT- nSSXv4n[R"XUS9n U $)z, Pad an image with zeros to the given size. Nr)fill)r{r)tvFpad) rYrrrNrOsize pad_bottom pad_rightpadding padded_images r _pad_imagesOwlv2ImageProcessor._pad_images sP RS) 6!] L a/wwv^D r!disable_groupingc [XS9upV0nUR5HupURU US9n XU'M [Xv5n U $)zx Unlike the Base class `self.pad` where all images are padded to the maximum image size, Owlv2 pads an image to square. r)r)ritemsrr ) rYrrrrQgrouped_imagesgrouped_images_indexprocessed_images_groupedr{stacked_imagesprocessed_imagess rrOwlv2ImageProcessor.padsg0EV/o,#% %3%9%9%; !E!---.N/=U + &<**BYr!imager anti_aliasingc >URUR4nURn[R"USS5R UR 5[R"U5R UR 5- nU(Ga#UcUS- S- RSS9nO[R"U5[R"U5-n[R"US:5(a [S5e[R"US:US:*-5(a[R"S5 [R"US:H5(aUn O[S[R"SU-5R5-S-n [ R""XSU S4UR%5S 9n OUn [&T U]QXS S 9$) a Resize an image as per the original implementation. Args: image (`Tensor`): Image to resize. size (`dict[str, int]`): Dictionary containing the height and width to resize the image to. anti_aliasing (`bool`, *optional*, defaults to `True`): Whether to apply anti-aliasing when downsampling the image. anti_aliasing_sigma (`float`, *optional*, defaults to `None`): Standard deviation for Gaussian kernel when downsampling the image. If `None`, it will be calculated automatically. r#Nrrr'zFAnti-aliasing standard deviation must be greater than or equal to zerozWAnti-aliasing standard deviation greater than zero but not down-sampling along all axesr)sigmaF)r antialias)rNrOr{rr<rBrCr* atleast_1d ones_likeanyrgwarningswarnceilrr gaussian_blurtolistrWresize) rYrrranti_aliasing_sigmarQ output_shape input_shapefactorsfiltered kernel_sizesr[s rrOwlv2ImageProcessor.resize/s, TZZ0 kk ,,{12/225<<@5<!77 E:N/= $U + &<**BDXY #xx(8_oxp /D /?0 ,"$%3%9%9%; !E>y $ .t _ 0=&u-&<((>@TU0E^fv/w,#% %3%9%9%; !E>!77~ZN/= $U + &<**BDXY.2B!CQ_``r!rU)ryN)rxg333333?N)rx)TN))__name__ __module__ __qualname____firstlineno__r BILINEARrr rr rrdefault_to_square crop_sizerdo_center_croprTrdo_convert_rgbmodel_input_namesrrrrrXrrr:r;rurrboolrr rstrrr__static_attributes__ __classcell__)r[s@rrLrLms !**H!JIC (DIINJLN'(N F# !5#8< 3-33!4;.5 3jKZ . % R` "!$  ^$ +   n   8# 4D4D4D 4D 4D4Dl7a^$7a7a 7a . 7a  7a7a7a7aDK'$.7a4;&-7a+7aj(4/7a 7a7ar!rL)$rtypingrr$torchvision.transforms.v2.functional transformsv2 functionalrimage_processing_backendsrimage_processing_utilsrimage_transformsrrr image_utilsr r r r processing_utilsrrutilsrrmodeling_owlv2rr r%r5rJrL__all__rUr!rrsy* 22;2__ZZ4/:G E  >na,nanab ! !r!