o W h6@sdddlmZddlZddlmZddlmZmZddlm Z m Z ddl m Z m Z GdddeZdS) )PathN)DetectionValidator)LOGGERops) OBBMetrics batch_probiou)output_to_rotated_target plot_imagescsfeZdZdZdfdd ZfddZddZd d Zd d Zd dZ ddZ ddZ ddZ Z S) OBBValidatora A class extending the DetectionValidator class for validation based on an Oriented Bounding Box (OBB) model. This validator specializes in evaluating models that predict rotated bounding boxes, commonly used for aerial and satellite imagery where objects can appear at various orientations. Attributes: args (dict): Configuration arguments for the validator. metrics (OBBMetrics): Metrics object for evaluating OBB model performance. is_dota (bool): Flag indicating whether the validation dataset is in DOTA format. Methods: init_metrics: Initialize evaluation metrics for YOLO. _process_batch: Process batch of detections and ground truth boxes to compute IoU matrix. _prepare_batch: Prepare batch data for OBB validation. _prepare_pred: Prepare predictions with scaled and padded bounding boxes. plot_predictions: Plot predicted bounding boxes on input images. pred_to_json: Serialize YOLO predictions to COCO json format. save_one_txt: Save YOLO detections to a txt file in normalized coordinates. eval_json: Evaluate YOLO output in JSON format and return performance statistics. Examples: >>> from ultralytics.models.yolo.obb import OBBValidator >>> args = dict(model="yolo11n-obb.pt", data="dota8.yaml") >>> validator = OBBValidator(args=args) >>> validator(model=args["model"]) Ncs0t|||||d|j_t|jdd|_dS)a Initialize OBBValidator and set task to 'obb', metrics to OBBMetrics. This constructor initializes an OBBValidator instance for validating Oriented Bounding Box (OBB) models. It extends the DetectionValidator class and configures it specifically for the OBB task. Args: dataloader (torch.utils.data.DataLoader, optional): Dataloader to be used for validation. save_dir (str | Path, optional): Directory to save results. pbar (bool, optional): Display progress bar during validation. args (dict, optional): Arguments containing validation parameters. _callbacks (list, optional): List of callback functions to be called during validation. obbT)save_dirplotN)super__init__argstaskrr metrics)self dataloaderr pbarr _callbacks __class__S/var/www/vscode/kcb/lib/python3.10/site-packages/ultralytics/models/yolo/obb/val.pyr*szOBBValidator.__init__cs6t||j|jjd}t|tod|v|_dS)z'Initialize evaluation metrics for YOLO.DOTAN) r init_metricsdatagetrsplit isinstancestris_dota)rmodelvalrrrr<s zOBBValidator.init_metricscCsRt|tj|ddddf|ddddfgdd}||dddf||S)ao Perform computation of the correct prediction matrix for a batch of detections and ground truth bounding boxes. Args: detections (torch.Tensor): A tensor of shape (N, 7) representing the detected bounding boxes and associated data. Each detection is represented as (x1, y1, x2, y2, conf, class, angle). gt_bboxes (torch.Tensor): A tensor of shape (M, 5) representing the ground truth bounding boxes. Each box is represented as (x1, y1, x2, y2, angle). gt_cls (torch.Tensor): A tensor of shape (M,) representing class labels for the ground truth bounding boxes. Returns: (torch.Tensor): The correct prediction matrix with shape (N, 10), which includes 10 IoU (Intersection over Union) levels for each detection, indicating the accuracy of predictions compared to the ground truth. Examples: >>> detections = torch.rand(100, 7) # 100 sample detections >>> gt_bboxes = torch.rand(50, 5) # 50 sample ground truth boxes >>> gt_cls = torch.randint(0, 5, (50,)) # 50 ground truth class labels >>> correct_matrix = OBBValidator._process_batch(detections, gt_bboxes, gt_cls) Note: This method relies on `batch_probiou` to calculate IoU between detections and ground truth bounding boxes. Ndim)rtorchcatmatch_predictions)r detections gt_bboxesgt_clsiourrr_process_batchBs8zOBBValidator._process_batchc Cs|d|k}|d|d}|d|}|d|}|djdd}|d |}t|rM|d dd ftj||jd gd tj||||dd|||||dS)a Prepare batch data for OBB validation with proper scaling and formatting. Args: si (int): Batch index to process. batch (dict): Dictionary containing batch data with keys: - batch_idx: Tensor of batch indices - cls: Tensor of class labels - bboxes: Tensor of bounding boxes - ori_shape: Original image shapes - img: Batch of images - ratio_pad: Ratio and padding information This method filters the batch data for a specific batch index, extracts class labels and bounding boxes, and scales the bounding boxes to the original image dimensions. batch_idxclsr'bboxes ori_shapeimgN ratio_pad.r&)device)rr;rTr9xywh)r4bboxr6imgszr9) squeezeshapelenmul_r+tensorr:r scale_boxes) rsibatchidxr4r>r6r?r9rrr_prepare_batch]s    *zOBBValidator._prepare_batchcCs<|}tj|d|ddddf|d|ddd|S)a8 Prepare predictions by scaling bounding boxes to original image dimensions. This method takes prediction tensors containing bounding box coordinates and scales them from the model's input dimensions to the original image dimensions using the provided batch information. Args: pred (torch.Tensor): Prediction tensor containing bounding box coordinates and other information. pbatch (dict): Dictionary containing batch information with keys: - imgsz (tuple): Model input image size. - ori_shape (tuple): Original image shape. - ratio_pad (tuple): Ratio and padding information for scaling. Returns: (torch.Tensor): Scaled prediction tensor with bounding boxes in original image dimensions. r?Nr&r6r9Tr<)clonerrE)rpredpbatchprednrrr _prepare_predys &zOBBValidator._prepare_predcCsFt|dgt||jjdR|d|jd|d|j|jddS)aU Plot predicted bounding boxes on input images and save the result. Args: batch (dict): Batch data containing images, file paths, and other metadata. preds (list): List of prediction tensors for each image in the batch. ni (int): Batch index used for naming the output file. Examples: >>> validator = OBBValidator() >>> batch = {"img": images, "im_file": paths} >>> preds = [torch.rand(10, 7)] # Example predictions for one image >>> validator.plot_predictions(batch, preds, 0) r7)max_detim_file val_batchz _pred.jpg)pathsfnamenameson_plotN)r rrrOr rTrU)rrGpredsnirrrplot_predictionss zOBBValidator.plot_predictionsc Cst|j}|r t|n|}tj|ddddf|ddddfgdd}t|dd}t t | | D]0\}\}} |j ||jt||dft||dfddd|Dd d| Dd qz-OBBValidator.pred_to_json..cSrZr[r]r_rrrrbrc)image_id category_idscorerboxpoly)rstem isnumericintr+r,rxywhr2xyxyxyxyview enumerateziptolistjdictappend class_mapitemr^) rrMfilenamerirdrgrhirbrrr pred_to_jsons 2"  zOBBValidator.pred_to_jsonc Csddl}ddlm}tj|ddddf|ddddfgdd}tj||ddddfgdd}||j|d|df|jd d|j|d j||d dS) a* Save YOLO OBB (Oriented Bounding Box) detections to a text file in normalized coordinates. Args: predn (torch.Tensor): Predicted detections with shape (N, 7) containing bounding boxes, confidence scores, class predictions, and angles in format (x, y, w, h, conf, cls, angle). save_conf (bool): Whether to save confidence scores in the text file. shape (tuple): Original image shape in format (height, width). file (Path | str): Output file path to save detections. Examples: >>> validator = OBBValidator() >>> predn = torch.tensor([[100, 100, 50, 30, 0.9, 0, 45]]) # One detection: x,y,w,h,conf,cls,angle >>> validator.save_one_txt(predn, True, (640, 480), "detection.txt") rN)Resultsr&r'r(r;)dtype)pathrTr ) save_conf) numpyultralytics.engine.resultsrzr+r,zerosuint8rTsave_txt) rrMr~rAfilenprzrboxesr rrr save_one_txts 2" zOBBValidator.save_one_txtcCs"|jjr|jrt|jrddl}ddl}ddlm}|j d}|j d}|j ddd| t |}t d|d |D]m}|d } |d } |j|d d dd} |d} t |d| dddd;} | | d| d| dd| d d| dd| dd| dd| dd| dd| ddWdn1swYq@|j d}|j ddd|t}t d|d |D]R}|d dd} |d }d!d"|||d dd#D\}}|d$|d |d d }} }|d|7<|d |7<|| |g|| |q|D]\} }t|}t|ddddfd}|ddddf|}|dddf}|ddddf}|ddddf|7<t||d%}||}t |ddddf!d&d'}tj"||ddddfgd&d(#D]u}|jt$|d&dd} d)d*|dd+D} t%|d+d} t |d| dddd;} | | d| d| dd| d d| dd| dd| dd| dd| dd| ddWdn 1swYqq |S),zHEvaluate YOLO output in JSON format and save predictions in DOTA format.rN) defaultdictzpredictions.jsonpredictions_txtT)parentsexist_okz'Saving predictions with DOTA format to z...rdrfrer; -rhTask1_z.txtazutf-8)encodingr8r\r&r*r{ predictions_merged_txtz.Saving merged predictions with DOTA format to __z \d+___\d+css|]}t|VqdS)N)rk)r`crrr sz)OBBValidator.eval_json..___rgg333333?r'rYr(cSrZr[r])r`rvrrrrbrcz*OBBValidator.eval_json..)&r save_jsonr#rBrqjsonre collectionsrr mkdirloadopenrinforTreplace writelineslistr compilefindallextendrritemsr+rDmaxrtrJr nms_rotatedrlrmr,rprkr^)rstatsrrr pred_jsonpred_txtrdrdrf classnamepfpred_merged_txtmerged_resultspatternrayr>r4max_whrscoresrxrvrrr eval_jsonsf    h  (  ""* hzOBBValidator.eval_json)NNNNN)__name__ __module__ __qualname____doc__rrr2rIrNrXryrr __classcell__rrrrr s r )pathlibrr+ultralytics.models.yolo.detectrultralytics.utilsrrultralytics.utils.metricsrrultralytics.utils.plottingrr r rrrrs