Fi{4ddlmZddlmZddlmZddlZddlZddl m cm Z ddl mZddlmZmZddlmZddlmZmZGd d eZy) ) annotations)Path)AnyN)DetectionValidator)LOGGERops)check_requirements)SegmentMetricsmask_iouceZdZdZddfd Zdfd Zdfd ZddZdfd Zdfd Z dfd Z dfd Z dfd Z dd Z dfd Zdfd Zdfd ZxZS)SegmentationValidatoraA class extending the DetectionValidator class for validation based on a segmentation model. This validator handles the evaluation of segmentation models, processing both bounding box and mask predictions to compute metrics such as mAP for both detection and segmentation tasks. Attributes: plot_masks (list): List to store masks for plotting. process (callable): Function to process masks based on save_json and save_txt flags. args (SimpleNamespace): Arguments for the validator. metrics (SegmentMetrics): Metrics calculator for segmentation tasks. stats (dict): Dictionary to store statistics during validation. Examples: >>> from ultralytics.models.yolo.segment import SegmentationValidator >>> args = dict(model="yolo26n-seg.pt", data="coco8-seg.yaml") >>> validator = SegmentationValidator(args=args) >>> validator() cxt|||||d|_d|j_t |_y)aInitialize SegmentationValidator and set task to 'segment', metrics to SegmentMetrics. Args: dataloader (torch.utils.data.DataLoader, optional): DataLoader to use for validation. save_dir (Path, optional): Directory to save results. args (dict, optional): Arguments for the validator. _callbacks (dict, optional): Dictionary of callback functions. Nsegment)super__init__processargstaskr metrics)self dataloadersave_dirr _callbacks __class__s e/var/www/pod-logistic/pod-ai/venv/lib/python3.12/site-packages/ultralytics/models/yolo/segment/val.pyrzSegmentationValidator.__init__&s4 XtZ@ " %' cRt||}|dj|d<|S)zPreprocess batch of images for YOLO segmentation validation. Args: batch (dict[str, Any]): Batch containing images and annotations. Returns: (dict[str, Any]): Preprocessed batch. masks)r preprocessfloat)rbatchrs rrz SegmentationValidator.preprocess4s/"5)w--/g rct|||jjr t d|jjs|jj rt j|_ yt j|_ y)zInitialize metrics and select mask processing function based on save_json flag. Args: model (torch.nn.Module): Model to validate. zfaster-coco-eval>=1.6.7N) r init_metricsr save_jsonr save_txtrprocess_mask_native process_maskr)rmodelrs rr#z"SegmentationValidator.init_metricsAsZ U# 99   8 926))2E2EI[I[s.. adaqaq rc ddzS)z5Return a formatted description of evaluation metrics.z,%22s%11s%11s%11s%11s%11s%11s%11s%11s%11s%11s) ClassImages InstanceszBox(PRmAP50 mAP50-95)zMask(Pr-r.r/)rs rget_desczSegmentationValidator.get_descMs$ )  rct|dtr|ddn|d}t| |d}|jddDcgc]}d|z }}t |D]\}}|j d}|jdr|j||||d|natjdg|jtjur|n|jddtj|dj |d <|Scc}w) aPost-process YOLO predictions and return output detections with proto. Args: preds (list[torch.Tensor]): Raw predictions from the model. Returns: (list[dict[str, torch.Tensor]]): Processed detection predictions with masks. rNextrabboxes)shape)dtypedevicer) isinstancetupler postprocessr8 enumeratepoprtorchzerosrr&uint8r:) rpredsprotoximgszipred coefficientrs rr=z!SegmentationValidator.postprocess]s *%(E:a a#E!H- % AB01Q00 ' GAt((7+K$$Q' U1X{DN% P[[a4<<33J3J#J%PUP[P[\]\^P_a++>00 M  1s Dc:t |||}|djd}|jjrR|d|}t j d|dz|jj|dd}||k(j}n|d|d|k(}|ru|dDcgc]%}|jtjur|n|dz'}}|jdd |k7r0tj|d |d d d}|jd }||d<|Scc}w)a:Prepare a batch for validation by processing images and targets. Args: si (int): Sample index within the batch. batch (dict[str, Any]): Batch data containing images and annotations. Returns: (dict[str, Any]): Prepared batch with processed annotations. clsrrr3)r: batch_idxrFr5NbilinearF)mode align_cornersg?)r_prepare_batchr8r overlap_maskr@aranger:viewr rrr&F interpolategt_) rsir!prepared_batchnlrindexs mask_sizers rrPz$SegmentationValidator._prepare_batchvs%/E: E " ( ( + 99 ! !'N2&ELLBF5<<@EEb!QOEe^**,E'N5#5#;>> preds = {"cls": torch.tensor([1, 0]), "masks": torch.rand(2, 640, 640), "bboxes": torch.rand(2, 4)} >>> batch = {"cls": torch.tensor([1, 0]), "masks": torch.rand(2, 640, 640), "bboxes": torch.rand(2, 4)} >>> correct_preds = validator._process_batch(preds, batch) Notes: - This method computes IoU between predicted and ground truth masks. - Overlapping masks are handled based on the overlap_mask argument setting. rKrr9rr3tp_m)r_process_batchr8nprAniouboolr flattenr match_predictionscpunumpyupdate)rrCr!tpgt_clsrciours rrdz$SegmentationValidator._process_batchs&W #E5 1u <<?a 5<#5#5a#8A#=88U5\//2DII>dKD5>11!4eGn6L6LQ6O6U6U6WXC))%,DHHJPPRD 64.! rc|D]}|d}|jd|jjkDr-tjd|jjdt j |d|jjt jj|d<t|)||||jjy)aPlot batch predictions with masks and bounding boxes. Args: batch (dict[str, Any]): Batch containing images and annotations. preds (list[dict[str, torch.Tensor]]): List of predictions from the model. ni (int): Batch index. rrz&Limiting validation plots to 'max_det=z' items.Nrb)max_det) r8rrqrwarningr@ as_tensorrBrjrplot_predictions)rr!rCniprrs rrtz&SegmentationValidator.plot_predictionss ^AgJE{{1~ 1 11!G HYHYGZZbcd/B1B1B)C5;;W[[]AgJ  ^  r499;L;L Mrc ddlm}|tj|d|dftjd|j t j|d|djd|d jdgd t j|d t j j|| y)aSave YOLO detections to a txt file in normalized coordinates in a specific format. Args: predn (dict[str, torch.Tensor]): Prediction dictionary containing 'bboxes', 'conf', 'cls', and 'masks' keys. save_conf (bool): Whether to save confidence scores. shape (tuple[int, int]): Shape of the original image. file (Path): File path to save the detections. r)Resultsr3rbNr7confrK)dimr)pathnamesboxesr) save_conf) ultralytics.engine.resultsrxrerArBr}r@cat unsqueezersr%)rprednrr8filerxs r save_one_txtz"SegmentationValidator.save_one_txts 7 HHeAha) :**))U8_eFm.E.Eb.I5QV.to_stringsF3v;' q Nq5VAE]++ADA!GA*+TAGaDT GAMM#a&) ,776? "rc>|ddddf|ddddfk7}tj|\}}|dz}g}t|jdD]}|||k(}t |rxtj |j }|jd|dj|jt |||djz nt ||g}||djdk(rdg|}|j||S)aIConvert multiple binary masks using Run-Length Encoding (RLE). Args: pixels (torch.Tensor): A 2D tensor where each row represents a flattened binary mask with shape [N, H*W]. Returns: (list[list[int]]): A list of RLE counts for each mask. Nr3rzr) r@whererr8rdifftolistinsertitemr)pixels transitionsrow_idxcol_idxrrG positionscounts r multi_encodez8SegmentationValidator.pred_to_json..multi_encodes!AB-6!SbS&>9K${{;7 GWkGF6<<?+ %#GqL1 y>!JJy188:ELLIaL$5$5$78LLVAY)B-2D2D2F!FG ^,E!9Q<$$&!+KKE e$ %Mrrr4r3rz)sizer segmentationN)r list[int]returnstr)rz torch.Tensorrr) transpose contiguousrSrr8rr pred_to_jsonr>jdict)rrpbatchrr pred_maskshwrrlesrrGrrs rrz"SegmentationValidator.pred_to_jsons #B @7^--a3>>@EEc%PW.FY[]^ W~##Aa(1j) BA KK!Q9Q<@ A B UF+dO ;DAq9:DJJD zA~ &~ 6 ;rcit|||dtj|dd|d|ddj iS)z.Scales predictions to the original image size.rN ori_shape ratio_pad)rr)r scale_predsr scale_masksbyte)rrrrs rrz!SegmentationValidator.scale_preds's\ g!%0 S__U7^D%96+;NZ`alZmndf   rc|jdz }|jddz |jrdnd|jjdz }t ||||ddgd d g S) z;Return COCO-style instance segmentation evaluation metrics.zpredictions.jsonr|rzinstances_val2017.jsonlvis_v1_z.jsonbboxsegmBoxMask)suffix)rdatais_cocorsplitr coco_evaluate)rstats pred_json anno_jsonrs r eval_jsonzSegmentationValidator.eval_json0syMM$66 IIf  +/<<'x GXX]=^ `  w$UIy66BR\aci[j$kkr)NNNN)rz dict | NonerNone)r!dict[str, Any]rr)r(ztorch.nn.Modulerr)rr)rCzlist[torch.Tensor]rlist[dict[str, torch.Tensor]])rWrr!rrr)rr)rCdict[str, torch.Tensor]r!rrzdict[str, np.ndarray])r!rrCrrurrr) rrrrgr8ztuple[int, int]rrrr)rrrrrr)rrrrrr)rrrr)__name__ __module__ __qualname____doc__rrr#r1r=rPr^rdrtrrrr __classcell__)rs@rr r sT& (  r 245 : N.&Q;f llrr ) __future__rpathlibrtypingrrkrer@torch.nn.functionalnn functionalrTultralytics.models.yolo.detectrultralytics.utilsrrultralytics.utils.checksr ultralytics.utils.metricsr r r r0rrrs9# =)7>fl.flr