^jdZddlmZddlZddlmZddlmZddlm Z m Z ddl m Z m Z dd lmZmZdd lmZmZerd d lmZdd lmZGddedZdddefdZddZeGddeZdgZy)z%Image processor class for SuperPoint.) TYPE_CHECKINGN)TorchvisionBackend) BatchFeature)group_images_by_shapereorder_images)PILImageResamplingSizeDict) ImagesKwargsUnpack) TensorTypeauto_docstring)#SuperPointKeypointDescriptionOutput) functionalceZdZUdZeed<y)SuperPointImageProcessorKwargsz do_grayscale (`bool`, *optional*, defaults to `self.do_grayscale`): Whether to convert the image to grayscale. Can be overridden by `do_grayscale` in the `preprocess` method. do_grayscaleN)__name__ __module__ __qualname____doc__bool__annotations__/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/superpoint/image_processing_superpoint.pyrr"s rrF)totalimage torch.Tensorreturnc $|jdks#|j|jdk(rdnddk(rytj|ddddddf|ddddddfk(xr.tj|ddddddf|ddddddfk(S)zAChecks if an image is grayscale (all RGB channels are identical).rrrT.N)ndimshapetorchallrs r is_grayscaler)+s zzA~%**/QqAQF 99U31a<(E#q!Q,,?? @ UYY c1aluS!Q\22FrcJt|r|Stj|dS)a Converts an image to grayscale format using the NTSC formula. Only support torch.Tensor. This function is supposed to return a 1-channel image, but it returns a 3-channel image with the same value in each channel, because of an issue that is discussed in : https://github.com/huggingface/transformers/pull/25786#issuecomment-1730176446 Args: image (torch.Tensor): The image to convert. r)num_output_channels)r)tvFrgb_to_grayscaler(s rconvert_to_grayscaler.4s$E   1 ==rceZdZeZej ZdddZdZ dZ dZ dZ dZ dZdeeffd Z dd ed d ed edddedededzdeezdzdedefdZdddeeezdeeed ffdZxZS)SuperPointImageProcessorii)heightwidthFTgp?Nkwargsc $t|di|y)Nr)super__init__)selfr3 __class__s rr6z!SuperPointImageProcessor.__init__Qs "6"rimagesr do_resizesizeresamplez7PILImageResampling | tvF.InterpolationMode | int | None do_rescalerescale_factordisable_groupingreturn_tensorsrr!c  t||\} } i} | jD]A\}}| r t|}|r|j|||}|r|j ||}|| |<Ct | | }t d|i|S)N)r?)r;r< pixel_values)data tensor_type)ritemsr.resizerescalerr)r7r9r:r;r<r=r>r?r@rr3grouped_imagesgrouped_images_indexprocessed_images_groupedr%stacked_imagesprocessed_imagess r _preprocessz$SuperPointImageProcessor._preprocessTs0EV^n/o,,#% %3%9%9%; = !E>!5n!E!%^$QY!Z!%nn!M.< $U + =**BDXY.2B!CQ_``routputsr target_sizesct|jt|k7r tdt|tr,t j ||jj}n|jddk7r td|}t j|dg}|j|dddfz}|jt j}g}t|j||j|jD]O\}}}} t j |j#d} || }|| }| | } |j%||| dQ|S)a Converts the raw output of [`SuperPointForKeypointDetection`] into lists of keypoints, scores and descriptors with coordinates absolute to the original image sizes. Args: outputs ([`SuperPointKeypointDescriptionOutput`]): Raw outputs of the model containing keypoints in a relative (x, y) format, with scores and descriptors. target_sizes (`torch.Tensor` or `list[tuple[int, int]]`): Tensor of shape `(batch_size, 2)` or list of tuples (`tuple[int, int]`) containing the target size `(height, width)` of each image in the batch. This must be the original image size (before any processing). Returns: `list[Dict]`: A list of dictionaries, each dictionary containing the keypoints in absolute format according to target_sizes, scores and descriptors for an image in the batch as predicted by the model. zRMake sure that you pass in as many target sizes as the batch dimension of the mask)devicerr#zTEach element of target_sizes must contain the size (h, w) of each image of the batchN) keypointsscores descriptors)lenmask ValueError isinstancelistr&tensorrQr%fliprRtoint32ziprSrTnonzerosqueezeappend) r7rNrO image_sizesmasked_keypointsresults image_maskrRrSrTindicess rpost_process_keypoint_detectionz8SuperPointImageProcessor.post_process_keypoint_detectionrsH$ w|| L 1 1qr r lD ),,|GLL do_normalizerr r6rYrr floatstrr rrMtupledictrg __classcell__)r8s@rr0r0Es1L!**HC (DIJNLL#(F!G##a^$aa a L a  aa+aj(4/aa ar{s}, ;2E74/H7\4>"[1[[| & &r