iFdZddlZddlmZddlZddlmcmcmZ ddl m Z ddl m Z mZddlmZmZmZmZmZddlmZmZd d lmZeGd d eZd gZy) z%Fast Image processor class for OWLv2.N)Optional) BatchFeature)group_images_by_shapereorder_images)OPENAI_CLIP_MEANOPENAI_CLIP_STDChannelDimensionPILImageResamplingSizeDict) TensorTypeauto_docstring)OwlViTImageProcessorFastc2eZdZejZeZeZ dddZ dZ dZ dZ dZdZdZdZdddded dfd Z ddedd edzded edfd Z dd ddeded dfdZdeddedededdededededeeezdzdeeezdzd edzdeezdzd efdZy)Owlv2ImageProcessorFasti)heightwidthgp?TNimagesz torch.Tensorconstant_valuereturnc|jdd\}}t||}||z }||z }dd||f}tj|||} | S)z< Pad an image with zeros to the given size. Nr)fill)shapemaxtvFpad) selfrrrrsize pad_bottom pad_rightpadding padded_images c/mnt/e/genesis-system/.venv/lib/python3.12/site-packages/transformers/models/owlv2/modular_owlv2.py _pad_imagesz#Owlv2ImageProcessorFast._pad_images6s\ RS) 65!F] 5L aJ/wwvw^D disable_groupingc t||\}}i}|jD]\}} |j| |} | ||<t||} | S)z Unlike the Base class `self.pad` where all images are padded to the maximum image size, Owlv2 pads an image to square. r()r)ritemsr&r) rrr(rkwargsgrouped_imagesgrouped_images_indexprocessed_images_groupedrstacked_imagesprocessed_imagess r%rzOwlv2ImageProcessorFast.padCsz0EV^n/o,,#% %3%9%9%; = !E>!---.N/= $U +  =**BDXYr'imager anti_aliasingc |j|jf}|j}tj|ddj |j tj|j |j z }|r||dz dz jd}ntj|tj|z}tj|dkr tdtj|dkD|dkzrtjdtj|dk(r|} n]dtjd|zjzdz} t!j"|| d| df|j% } n|} t!j&| |j|jfd } | S) az 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. rNr)minzFAnti-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)rrrtorchtensortodeviceclamp atleast_1d ones_likeany ValueErrorwarningswarnceilintr gaussian_blurtolistresize) rr2r r3anti_aliasing_sigmar, output_shape input_shapefactorsfiltered kernel_sizesouts r%rHzOwlv2ImageProcessorFast.resize[s, TZZ0 kk ,,{12/225<<@5<!77 NE:yN/= $U +  =**BDXY #xx(8_oxp /D /?0 ,,"$%3%9%9%; > !E> $ ("/&6&<&< !,! 1>&u- >((>@TU0E^fv/w,,#% %3%9%9%; = !E>!77~|ZQZN/= $U +  =**BDXY.2B!CQ_``r')rZ)TN)__name__ __module__ __qualname__r BILINEARresamplerrVr rWr rTrPrSrUrR crop_sizedo_center_cropfloatr&listboolrr rHrstrr rrdr'r%rr(s!**H!JIC (DNIJL FIN . % R` "!$  ^$ +   n   8# 666 6 6p<a^$<a<a <a  78 <a  <a<a<a<aDK'$.<a4;&-<a+<aj(4/<a <ar'r)__doc__rBtypingrr9$torchvision.transforms.v2.functional transformsv2 functionalrimage_processing_utils_fastrimage_transformsrr image_utilsrr r r r utilsr r#owlvit.image_processing_owlvit_fastrr__all__rpr'r%r}sf, 22FKfa6fafaR % %r'