o  h%@sddlmZmZmZmZddlZddlmZm Z m Z m Z m Z ddl mZmZe r5ddlmZddlmZe rDdd lmZmZmZmZe eZeeefZeeZeed d Gd d d eZdS))AnyDictListUnionN)add_end_docstringsis_torch_availableis_vision_availableloggingrequires_backends)Pipelinebuild_pipeline_init_args)Image) load_image)*MODEL_FOR_IMAGE_SEGMENTATION_MAPPING_NAMES-MODEL_FOR_INSTANCE_SEGMENTATION_MAPPING_NAMES-MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES.MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPING_NAMEST)has_image_processorcsfeZdZdZfddZddZddeeee fffdd Z dd d Z d d Z dddZ ZS)ImageSegmentationPipelinea Image segmentation pipeline using any `AutoModelForXXXSegmentation`. This pipeline predicts masks of objects and their classes. Example: ```python >>> from transformers import pipeline >>> segmenter = pipeline(model="facebook/detr-resnet-50-panoptic") >>> segments = segmenter("https://huggingface.co/datasets/Narsil/image_dummy/raw/main/parrots.png") >>> len(segments) 2 >>> segments[0]["label"] 'bird' >>> segments[1]["label"] 'bird' >>> type(segments[0]["mask"]) # This is a black and white mask showing where is the bird on the original image. >>> segments[0]["mask"].size (768, 512) ``` This image segmentation pipeline can currently be loaded from [`pipeline`] using the following task identifier: `"image-segmentation"`. See the list of available models on [huggingface.co/models](https://huggingface.co/models?filter=image-segmentation). csltj|i||jdkrtd|jdt|dt}|t |t |t | |dS)NtfzThe z is only available in PyTorch.vision) super__init__ framework ValueError __class__r rcopyupdaterrrcheck_model_type)selfargskwargsmappingr}/var/www/html/construction_image-detection-poc/venv/lib/python3.10/site-packages/transformers/pipelines/image_segmentation.pyrCs     z"ImageSegmentationPipeline.__init__cKsi}i}d|vr|d|d<|d|d<d|vr|d|d<d|vr(|d|d<d|vr2|d|d<d|vr<|d|d<|i|fS)Nsubtask thresholdmask_thresholdoverlap_mask_area_thresholdtimeoutr&)r!r#preprocess_kwargspostprocess_kwargsr&r&r'_sanitize_parametersPs       z.ImageSegmentationPipeline._sanitize_parametersNreturnc s6d|vr |d}|durtdtj|fi|S)a Perform segmentation (detect masks & classes) in the image(s) passed as inputs. Args: inputs (`str`, `List[str]`, `PIL.Image` or `List[PIL.Image]`): The pipeline handles three types of images: - A string containing an HTTP(S) link pointing to an image - A string containing a local path to an image - An image loaded in PIL directly The pipeline accepts either a single image or a batch of images. Images in a batch must all be in the same format: all as HTTP(S) links, all as local paths, or all as PIL images. subtask (`str`, *optional*): Segmentation task to be performed, choose [`semantic`, `instance` and `panoptic`] depending on model capabilities. If not set, the pipeline will attempt tp resolve in the following order: `panoptic`, `instance`, `semantic`. threshold (`float`, *optional*, defaults to 0.9): Probability threshold to filter out predicted masks. mask_threshold (`float`, *optional*, defaults to 0.5): Threshold to use when turning the predicted masks into binary values. overlap_mask_area_threshold (`float`, *optional*, defaults to 0.5): Mask overlap threshold to eliminate small, disconnected segments. timeout (`float`, *optional*, defaults to None): The maximum time in seconds to wait for fetching images from the web. If None, no timeout is set and the call may block forever. Return: A dictionary or a list of dictionaries containing the result. If the input is a single image, will return a list of dictionaries, if the input is a list of several images, will return a list of list of dictionaries corresponding to each image. The dictionaries contain the mask, label and score (where applicable) of each detected object and contains the following keys: - **label** (`str`) -- The class label identified by the model. - **mask** (`PIL.Image`) -- A binary mask of the detected object as a Pil Image of shape (width, height) of the original image. Returns a mask filled with zeros if no object is found. - **score** (*optional* `float`) -- Optionally, when the model is capable of estimating a confidence of the "object" described by the label and the mask. imagesNzICannot call the image-classification pipeline without an inputs argument!)poprr__call__)r!inputsr#r%r&r'r3as + z"ImageSegmentationPipeline.__call__cCst||d}|j|jfg}|jjjjdkrL|duri}nd|gi}|jd |gdd|}|jdkr8| |j }|j |dd|jjj |jdd|d<n|j|gdd}|jdkr_| |j }||d <|S) N)r,OneFormerConfig task_inputspt)r1return_tensors max_length)paddingr9r8 input_ids target_sizer&) rheightwidthmodelconfigr__name__image_processorrto torch_dtype tokenizer task_seq_len)r!imager(r,r<r#r4r&r&r' preprocesss,       z$ImageSegmentationPipeline.preprocesscCs&|d}|jdi|}||d<|S)Nr<r&)r2r?)r! model_inputsr< model_outputsr&r&r'_forwards z"ImageSegmentationPipeline._forward??cCsd}|dvrt|jdr|jj}n|dvrt|jdr|jj}|durj||||||ddd}g}|d} |d D]-} | | d kd } tj| tj d d } |j j j | d} | d} | | | | dq:|S|dvrt|jdr|jj||ddd}g}|} t| }|D]#} | | kd } tj| tj d d } |j j j | } | d| | dq|Std|dt|j )N>Npanoptic"post_process_panoptic_segmentation>Ninstance"post_process_instance_segmentationr<)r)r*r+ target_sizesr segmentation segments_infoidL)modelabel_idscore)rZlabelmask>Nsemantic"post_process_semantic_segmentation)rRzSubtask z is not supported for model )hasattrrBrOrQr fromarraynumpyastypenpuint8r?r@id2labelappendr^uniquertype)r!rJr(r)r*r+fnoutputs annotationrSsegmentr\r[rZlabelsr&r&r' postprocesssP    z%ImageSegmentationPipeline.postprocess)N)NN)NrLrMrM)rA __module__ __qualname____doc__rr/r Predictionsr Predictionr3rHrKrn __classcell__r&r&r%r'rs #  1r) typingrrrrrarcutilsrrr r r baser rPILr image_utilsrmodels.auto.modeling_autorrrr get_loggerrAloggerstrrsrrrr&r&r&r's