o oi@spddlmZddlmZddlmZddlmZddlm Z ddl m Z eGdddZ eGd d d Z d S) ) annotations) dataclass)Optional)Tensor) KORNIA_CHECK)resizec@sReZdZUdZded<ded<dZded<edd d ZdddZddddZ dS)SegmentationResultsa`Encapsulate the results obtained by a Segmentation model. Args: logits: Results logits with shape :math:`(B, C, H, W)`, where :math:`C` refers to the number of predicted masks scores: The scores from the logits. Shape :math:`(B, C)` mask_threshold: The threshold value to generate the `binary_masks` from the `logits` rlogitsscoresgfloatmask_thresholdreturncCs"|jdur |j}n|j}||jkS)aBinary mask generated from logits considering the mask_threshold. Shape will be the same of logits :math:`(B, C, H, W)` where :math:`C` is the number masks predicted. .. note:: If you run `original_res_logits`, this will generate the masks based on the original resolution logits. Otherwise, this will use the low resolution logits (self.logits). N)_original_res_logitsr r )selfxrT/home/ubuntu/.local/lib/python3.10/site-packages/kornia/contrib/models/structures.py binary_masks+s  z SegmentationResults.binary_masks input_sizetuple[int, int] original_sizeimage_size_encoderOptional[tuple[int, int]]cCs^|j}t|trt||dddd}|dd|dd|df}t||dddd}||_|jS)aaRemove padding and upscale the logits to the original image size. Resize to image encoder input -> remove padding (bottom and right) -> Resize to original size .. note:: This method set a internal `original_res_logits` which will be used if available for the binary masks. Args: input_size: The size of the image input to the model, in (H, W) format. Used to remove padding. original_size: The original size of the image before resizing for input to the model, in (H, W) format. image_size_encoder: The size of the input image for image encoder, in (H, W) format. Used to resize the logits back to encoder resolution before remove the padding. Returns: Batched logits in :math:`(K, C, H, W)` format, where (H, W) is given by original_size. bilinearF)size interpolation align_corners antialias.Nr)r isinstancetuplerr)rrrrrrrroriginal_res_logits<s z'SegmentationResults.original_res_logitsrdimintcCs:|j||_|j||_t|jtr|j||_|S)z7Realize a squeeze for the dim given for all properties.)r squeezer rrr)rr"rrrr$Zs  zSegmentationResults.squeezeN)r r)rrrrrrr r)r)r"r#r r) __name__ __module__ __qualname____doc____annotations__r propertyrr!r$rrrrrs     rc@s\eZdZUdZdZded<dZded<dZded<dd d Ze dd d Z e dddZ dS)PromptsaEncapsulate the prompts inputs for a Model. Args: points: A tuple with the keypoints (coordinates x, y) and their respective labels. Shape :math:`(K, N, 2)` for the keypoints, and :math:`(K, N)` boxes: Batched box inputs, with shape :math:`(K, 4)`. Expected to be into xyxy format. masks: Batched mask prompts to the model with shape :math:`(K, 1, H, W)` NzOptional[tuple[Tensor, Tensor]]pointsOptional[Tensor]boxesmasksr NonecCsBt|jtrt|jtrt|jjd|jjdkddSdSdS)Nrz,The prompts should have the same batch size!)r keypointsrr.rshaperrrr __post_init__ts"zPrompts.__post_init__cCt|jtr |jdSdS)z The keypoints from the `points`.rNrr,r r3rrrr1xzPrompts.keypointscCr5)z'The keypoints labels from the `points`.rNr6r3rrrkeypoints_labels}r7zPrompts.keypoints_labels)r r0)r r-) r%r&r'r(r,r)r.r/r4r*r1r8rrrrr+ds     r+N) __future__r dataclassesrtypingr kornia.corerkornia.core.checkrkornia.geometry.transformrrr+rrrrs      G