v1.0

ultralytics/models/yolo/classify/train.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from copy import copy
+
+import torch
+
+from ultralytics.data import ClassificationDataset, build_dataloader
+from ultralytics.engine.trainer import BaseTrainer
+from ultralytics.models import yolo
+from ultralytics.nn.tasks import ClassificationModel
+from ultralytics.utils import DEFAULT_CFG, LOGGER, RANK
+from ultralytics.utils.plotting import plot_images, plot_results
+from ultralytics.utils.torch_utils import is_parallel, strip_optimizer, torch_distributed_zero_first
+
+
+class ClassificationTrainer(BaseTrainer):
+    """
+    A class extending the BaseTrainer class for training based on a classification model.
+
+    Notes:
+        - Torchvision classification models can also be passed to the 'model' argument, i.e. model='resnet18'.
+
+    Example:
+        ```python
+        from ultralytics.models.yolo.classify import ClassificationTrainer
+
+        args = dict(model="yolov8n-cls.pt", data="imagenet10", epochs=3)
+        trainer = ClassificationTrainer(overrides=args)
+        trainer.train()
+        ```
+    """
+
+    def __init__(self, cfg=DEFAULT_CFG, overrides=None, _callbacks=None):
+        """Initialize a ClassificationTrainer object with optional configuration overrides and callbacks."""
+        if overrides is None:
+            overrides = {}
+        overrides["task"] = "classify"
+        if overrides.get("imgsz") is None:
+            overrides["imgsz"] = 224
+        super().__init__(cfg, overrides, _callbacks)
+
+    def set_model_attributes(self):
+        """Set the YOLO model's class names from the loaded dataset."""
+        self.model.names = self.data["names"]
+
+    def get_model(self, cfg=None, weights=None, verbose=True):
+        """Returns a modified PyTorch model configured for training YOLO."""
+        model = ClassificationModel(cfg, nc=self.data["nc"], verbose=verbose and RANK == -1)
+        if weights:
+            model.load(weights)
+
+        for m in model.modules():
+            if not self.args.pretrained and hasattr(m, "reset_parameters"):
+                m.reset_parameters()
+            if isinstance(m, torch.nn.Dropout) and self.args.dropout:
+                m.p = self.args.dropout  # set dropout
+        for p in model.parameters():
+            p.requires_grad = True  # for training
+        return model
+
+    def setup_model(self):
+        """Load, create or download model for any task."""
+        import torchvision  # scope for faster 'import ultralytics'
+
+        if str(self.model) in torchvision.models.__dict__:
+            self.model = torchvision.models.__dict__[self.model](
+                weights="IMAGENET1K_V1" if self.args.pretrained else None
+            )
+            ckpt = None
+        else:
+            ckpt = super().setup_model()
+        ClassificationModel.reshape_outputs(self.model, self.data["nc"])
+        return ckpt
+
+    def build_dataset(self, img_path, mode="train", batch=None):
+        """Creates a ClassificationDataset instance given an image path, and mode (train/test etc.)."""
+        return ClassificationDataset(root=img_path, args=self.args, augment=mode == "train", prefix=mode)
+
+    def get_dataloader(self, dataset_path, batch_size=16, rank=0, mode="train"):
+        """Returns PyTorch DataLoader with transforms to preprocess images for inference."""
+        with torch_distributed_zero_first(rank):  # init dataset *.cache only once if DDP
+            dataset = self.build_dataset(dataset_path, mode)
+
+        loader = build_dataloader(dataset, batch_size, self.args.workers, rank=rank)
+        # Attach inference transforms
+        if mode != "train":
+            if is_parallel(self.model):
+                self.model.module.transforms = loader.dataset.torch_transforms
+            else:
+                self.model.transforms = loader.dataset.torch_transforms
+        return loader
+
+    def preprocess_batch(self, batch):
+        """Preprocesses a batch of images and classes."""
+        batch["img"] = batch["img"].to(self.device)
+        batch["cls"] = batch["cls"].to(self.device)
+        return batch
+
+    def progress_string(self):
+        """Returns a formatted string showing training progress."""
+        return ("\n" + "%11s" * (4 + len(self.loss_names))) % (
+            "Epoch",
+            "GPU_mem",
+            *self.loss_names,
+            "Instances",
+            "Size",
+        )
+
+    def get_validator(self):
+        """Returns an instance of ClassificationValidator for validation."""
+        self.loss_names = ["loss"]
+        return yolo.classify.ClassificationValidator(
+            self.test_loader, self.save_dir, args=copy(self.args), _callbacks=self.callbacks
+        )
+
+    def label_loss_items(self, loss_items=None, prefix="train"):
+        """
+        Returns a loss dict with labelled training loss items tensor.
+
+        Not needed for classification but necessary for segmentation & detection
+        """
+        keys = [f"{prefix}/{x}" for x in self.loss_names]
+        if loss_items is None:
+            return keys
+        loss_items = [round(float(loss_items), 5)]
+        return dict(zip(keys, loss_items))
+
+    def plot_metrics(self):
+        """Plots metrics from a CSV file."""
+        plot_results(file=self.csv, classify=True, on_plot=self.on_plot)  # save results.png
+
+    def final_eval(self):
+        """Evaluate trained model and save validation results."""
+        for f in self.last, self.best:
+            if f.exists():
+                strip_optimizer(f)  # strip optimizers
+                if f is self.best:
+                    LOGGER.info(f"\nValidating {f}...")
+                    self.validator.args.data = self.args.data
+                    self.validator.args.plots = self.args.plots
+                    self.metrics = self.validator(model=f)
+                    self.metrics.pop("fitness", None)
+                    self.run_callbacks("on_fit_epoch_end")
+
+    def plot_training_samples(self, batch, ni):
+        """Plots training samples with their annotations."""
+        plot_images(
+            images=batch["img"],
+            batch_idx=torch.arange(len(batch["img"])),
+            cls=batch["cls"].view(-1),  # warning: use .view(), not .squeeze() for Classify models
+            fname=self.save_dir / f"train_batch{ni}.jpg",
+            on_plot=self.on_plot,
+        )

ultralytics/models/yolo/classify/val.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+import torch
+
+from ultralytics.data import ClassificationDataset, build_dataloader
+from ultralytics.engine.validator import BaseValidator
+from ultralytics.utils import LOGGER
+from ultralytics.utils.metrics import ClassifyMetrics, ConfusionMatrix
+from ultralytics.utils.plotting import plot_images
+
+
+class ClassificationValidator(BaseValidator):
+    """
+    A class extending the BaseValidator class for validation based on a classification model.
+
+    Notes:
+        - Torchvision classification models can also be passed to the 'model' argument, i.e. model='resnet18'.
+
+    Example:
+        ```python
+        from ultralytics.models.yolo.classify import ClassificationValidator
+
+        args = dict(model="yolov8n-cls.pt", data="imagenet10")
+        validator = ClassificationValidator(args=args)
+        validator()
+        ```
+    """
+
+    def __init__(self, dataloader=None, save_dir=None, pbar=None, args=None, _callbacks=None):
+        """Initializes ClassificationValidator instance with args, dataloader, save_dir, and progress bar."""
+        super().__init__(dataloader, save_dir, pbar, args, _callbacks)
+        self.targets = None
+        self.pred = None
+        self.args.task = "classify"
+        self.metrics = ClassifyMetrics()
+
+    def get_desc(self):
+        """Returns a formatted string summarizing classification metrics."""
+        return ("%22s" + "%11s" * 2) % ("classes", "top1_acc", "top5_acc")
+
+    def init_metrics(self, model):
+        """Initialize confusion matrix, class names, and top-1 and top-5 accuracy."""
+        self.names = model.names
+        self.nc = len(model.names)
+        self.confusion_matrix = ConfusionMatrix(nc=self.nc, conf=self.args.conf, task="classify")
+        self.pred = []
+        self.targets = []
+
+    def preprocess(self, batch):
+        """Preprocesses input batch and returns it."""
+        batch["img"] = batch["img"].to(self.device, non_blocking=True)
+        batch["img"] = batch["img"].half() if self.args.half else batch["img"].float()
+        batch["cls"] = batch["cls"].to(self.device)
+        return batch
+
+    def update_metrics(self, preds, batch):
+        """Updates running metrics with model predictions and batch targets."""
+        n5 = min(len(self.names), 5)
+        self.pred.append(preds.argsort(1, descending=True)[:, :n5].type(torch.int32).cpu())
+        self.targets.append(batch["cls"].type(torch.int32).cpu())
+
+    def finalize_metrics(self, *args, **kwargs):
+        """Finalizes metrics of the model such as confusion_matrix and speed."""
+        self.confusion_matrix.process_cls_preds(self.pred, self.targets)
+        if self.args.plots:
+            for normalize in True, False:
+                self.confusion_matrix.plot(
+                    save_dir=self.save_dir, names=self.names.values(), normalize=normalize, on_plot=self.on_plot
+                )
+        self.metrics.speed = self.speed
+        self.metrics.confusion_matrix = self.confusion_matrix
+        self.metrics.save_dir = self.save_dir
+
+    def postprocess(self, preds):
+        """Preprocesses the classification predictions."""
+        return preds[0] if isinstance(preds, (list, tuple)) else preds
+
+    def get_stats(self):
+        """Returns a dictionary of metrics obtained by processing targets and predictions."""
+        self.metrics.process(self.targets, self.pred)
+        return self.metrics.results_dict
+
+    def build_dataset(self, img_path):
+        """Creates and returns a ClassificationDataset instance using given image path and preprocessing parameters."""
+        return ClassificationDataset(root=img_path, args=self.args, augment=False, prefix=self.args.split)
+
+    def get_dataloader(self, dataset_path, batch_size):
+        """Builds and returns a data loader for classification tasks with given parameters."""
+        dataset = self.build_dataset(dataset_path)
+        return build_dataloader(dataset, batch_size, self.args.workers, rank=-1)
+
+    def print_results(self):
+        """Prints evaluation metrics for YOLO object detection model."""
+        pf = "%22s" + "%11.3g" * len(self.metrics.keys)  # print format
+        LOGGER.info(pf % ("all", self.metrics.top1, self.metrics.top5))
+
+    def plot_val_samples(self, batch, ni):
+        """Plot validation image samples."""
+        plot_images(
+            images=batch["img"],
+            batch_idx=torch.arange(len(batch["img"])),
+            cls=batch["cls"].view(-1),  # warning: use .view(), not .squeeze() for Classify models
+            fname=self.save_dir / f"val_batch{ni}_labels.jpg",
+            names=self.names,
+            on_plot=self.on_plot,
+        )
+
+    def plot_predictions(self, batch, preds, ni):
+        """Plots predicted bounding boxes on input images and saves the result."""
+        plot_images(
+            batch["img"],
+            batch_idx=torch.arange(len(batch["img"])),
+            cls=torch.argmax(preds, dim=1),
+            fname=self.save_dir / f"val_batch{ni}_pred.jpg",
+            names=self.names,
+            on_plot=self.on_plot,
+        )  # pred

ultralytics/models/yolo/detect/__init__.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from .predict import DetectionPredictor
+from .train import DetectionTrainer
+from .val import DetectionValidator
+
+__all__ = "DetectionPredictor", "DetectionTrainer", "DetectionValidator"

ultralytics/models/yolo/detect/predict.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from ultralytics.engine.predictor import BasePredictor
+from ultralytics.engine.results import Results
+from ultralytics.utils import ops
+
+
+class DetectionPredictor(BasePredictor):
+    """
+    A class extending the BasePredictor class for prediction based on a detection model.
+
+    Example:
+        ```python
+        from ultralytics.utils import ASSETS
+        from ultralytics.models.yolo.detect import DetectionPredictor
+
+        args = dict(model="yolo11n.pt", source=ASSETS)
+        predictor = DetectionPredictor(overrides=args)
+        predictor.predict_cli()
+        ```
+    """
+
+    def postprocess(self, preds, img, orig_imgs):
+        """Post-processes predictions and returns a list of Results objects."""
+        preds = ops.non_max_suppression(
+            preds,
+            self.args.conf,
+            self.args.iou,
+            agnostic=self.args.agnostic_nms,
+            max_det=self.args.max_det,
+            classes=self.args.classes,
+        )
+
+        if not isinstance(orig_imgs, list):  # input images are a torch.Tensor, not a list
+            orig_imgs = ops.convert_torch2numpy_batch(orig_imgs)
+
+        results = []
+        for pred, orig_img, img_path in zip(preds, orig_imgs, self.batch[0]):
+            pred[:, :4] = ops.scale_boxes(img.shape[2:], pred[:, :4], orig_img.shape)
+            results.append(Results(orig_img, path=img_path, names=self.model.names, boxes=pred))
+        return results

ultralytics/models/yolo/detect/train.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+import math
+import random
+from copy import copy
+
+import numpy as np
+import torch.nn as nn
+
+from ultralytics.data import build_dataloader, build_yolo_dataset
+from ultralytics.engine.trainer import BaseTrainer
+from ultralytics.models import yolo
+from ultralytics.nn.tasks import DetectionModel
+from ultralytics.utils import LOGGER, RANK
+from ultralytics.utils.plotting import plot_images, plot_labels, plot_results
+from ultralytics.utils.torch_utils import de_parallel, torch_distributed_zero_first
+
+
+class DetectionTrainer(BaseTrainer):
+    """
+    A class extending the BaseTrainer class for training based on a detection model.
+
+    Example:
+        ```python
+        from ultralytics.models.yolo.detect import DetectionTrainer
+
+        args = dict(model="yolo11n.pt", data="coco8.yaml", epochs=3)
+        trainer = DetectionTrainer(overrides=args)
+        trainer.train()
+        ```
+    """
+
+    def build_dataset(self, img_path, mode="train", batch=None):
+        """
+        Build YOLO Dataset.
+
+        Args:
+            img_path (str): Path to the folder containing images.
+            mode (str): `train` mode or `val` mode, users are able to customize different augmentations for each mode.
+            batch (int, optional): Size of batches, this is for `rect`. Defaults to None.
+        """
+        gs = max(int(de_parallel(self.model).stride.max() if self.model else 0), 32)
+        return build_yolo_dataset(self.args, img_path, batch, self.data, mode=mode, rect=mode == "val", stride=gs)
+
+    def get_dataloader(self, dataset_path, batch_size=16, rank=0, mode="train"):
+        """Construct and return dataloader."""
+        assert mode in {"train", "val"}, f"Mode must be 'train' or 'val', not {mode}."
+        with torch_distributed_zero_first(rank):  # init dataset *.cache only once if DDP
+            dataset = self.build_dataset(dataset_path, mode, batch_size)
+        shuffle = mode == "train"
+        if getattr(dataset, "rect", False) and shuffle:
+            LOGGER.warning("WARNING ⚠️ 'rect=True' is incompatible with DataLoader shuffle, setting shuffle=False")
+            shuffle = False
+        workers = self.args.workers if mode == "train" else self.args.workers * 2
+        return build_dataloader(dataset, batch_size, workers, shuffle, rank)  # return dataloader
+
+    def preprocess_batch(self, batch):
+        """Preprocesses a batch of images by scaling and converting to float."""
+        batch["img"] = batch["img"].to(self.device, non_blocking=True).float() / 255
+        if self.args.multi_scale:
+            imgs = batch["img"]
+            sz = (
+                random.randrange(int(self.args.imgsz * 0.5), int(self.args.imgsz * 1.5 + self.stride))
+                // self.stride
+                * self.stride
+            )  # size
+            sf = sz / max(imgs.shape[2:])  # scale factor
+            if sf != 1:
+                ns = [
+                    math.ceil(x * sf / self.stride) * self.stride for x in imgs.shape[2:]
+                ]  # new shape (stretched to gs-multiple)
+                imgs = nn.functional.interpolate(imgs, size=ns, mode="bilinear", align_corners=False)
+            batch["img"] = imgs
+        return batch
+
+    def set_model_attributes(self):
+        """Nl = de_parallel(self.model).model[-1].nl  # number of detection layers (to scale hyps)."""
+        # self.args.box *= 3 / nl  # scale to layers
+        # self.args.cls *= self.data["nc"] / 80 * 3 / nl  # scale to classes and layers
+        # self.args.cls *= (self.args.imgsz / 640) ** 2 * 3 / nl  # scale to image size and layers
+        self.model.nc = self.data["nc"]  # attach number of classes to model
+        self.model.names = self.data["names"]  # attach class names to model
+        self.model.args = self.args  # attach hyperparameters to model
+        # TODO: self.model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc
+
+    def get_model(self, cfg=None, weights=None, verbose=True):
+        """Return a YOLO detection model."""
+        model = DetectionModel(cfg, nc=self.data["nc"], verbose=verbose and RANK == -1)
+        if weights:
+            model.load(weights)
+        return model
+
+    def get_validator(self):
+        """Returns a DetectionValidator for YOLO model validation."""
+        self.loss_names = "box_loss", "cls_loss", "dfl_loss"
+        return yolo.detect.DetectionValidator(
+            self.test_loader, save_dir=self.save_dir, args=copy(self.args), _callbacks=self.callbacks
+        )
+
+    def label_loss_items(self, loss_items=None, prefix="train"):
+        """
+        Returns a loss dict with labelled training loss items tensor.
+
+        Not needed for classification but necessary for segmentation & detection
+        """
+        keys = [f"{prefix}/{x}" for x in self.loss_names]
+        if loss_items is not None:
+            loss_items = [round(float(x), 5) for x in loss_items]  # convert tensors to 5 decimal place floats
+            return dict(zip(keys, loss_items))
+        else:
+            return keys
+
+    def progress_string(self):
+        """Returns a formatted string of training progress with epoch, GPU memory, loss, instances and size."""
+        return ("\n" + "%11s" * (4 + len(self.loss_names))) % (
+            "Epoch",
+            "GPU_mem",
+            *self.loss_names,
+            "Instances",
+            "Size",
+        )
+
+    def plot_training_samples(self, batch, ni):
+        """Plots training samples with their annotations."""
+        plot_images(
+            images=batch["img"],
+            batch_idx=batch["batch_idx"],
+            cls=batch["cls"].squeeze(-1),
+            bboxes=batch["bboxes"],
+            paths=batch["im_file"],
+            fname=self.save_dir / f"train_batch{ni}.jpg",
+            on_plot=self.on_plot,
+        )
+
+    def plot_metrics(self):
+        """Plots metrics from a CSV file."""
+        plot_results(file=self.csv, on_plot=self.on_plot)  # save results.png
+
+    def plot_training_labels(self):
+        """Create a labeled training plot of the YOLO model."""
+        boxes = np.concatenate([lb["bboxes"] for lb in self.train_loader.dataset.labels], 0)
+        cls = np.concatenate([lb["cls"] for lb in self.train_loader.dataset.labels], 0)
+        plot_labels(boxes, cls.squeeze(), names=self.data["names"], save_dir=self.save_dir, on_plot=self.on_plot)
+
+    def auto_batch(self):
+        """Get batch size by calculating memory occupation of model."""
+        train_dataset = self.build_dataset(self.trainset, mode="train", batch=16)
+        # 4 for mosaic augmentation
+        max_num_obj = max(len(label["cls"]) for label in train_dataset.labels) * 4
+        return super().auto_batch(max_num_obj)

ultralytics/models/yolo/detect/val.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+import os
+from pathlib import Path
+
+import numpy as np
+import torch
+
+from ultralytics.data import build_dataloader, build_yolo_dataset, converter
+from ultralytics.engine.validator import BaseValidator
+from ultralytics.utils import LOGGER, ops
+from ultralytics.utils.checks import check_requirements
+from ultralytics.utils.metrics import ConfusionMatrix, DetMetrics, box_iou
+from ultralytics.utils.plotting import output_to_target, plot_images
+
+
+class DetectionValidator(BaseValidator):
+    """
+    A class extending the BaseValidator class for validation based on a detection model.
+
+    Example:
+        ```python
+        from ultralytics.models.yolo.detect import DetectionValidator
+
+        args = dict(model="yolo11n.pt", data="coco8.yaml")
+        validator = DetectionValidator(args=args)
+        validator()
+        ```
+    """
+
+    def __init__(self, dataloader=None, save_dir=None, pbar=None, args=None, _callbacks=None):
+        """Initialize detection model with necessary variables and settings."""
+        super().__init__(dataloader, save_dir, pbar, args, _callbacks)
+        self.nt_per_class = None
+        self.nt_per_image = None
+        self.is_coco = False
+        self.is_lvis = False
+        self.class_map = None
+        self.args.task = "detect"
+        self.metrics = DetMetrics(save_dir=self.save_dir, on_plot=self.on_plot)
+        self.iouv = torch.linspace(0.5, 0.95, 10)  # IoU vector for mAP@0.5:0.95
+        self.niou = self.iouv.numel()
+        self.lb = []  # for autolabelling
+        if self.args.save_hybrid:
+            LOGGER.warning(
+                "WARNING ⚠️ 'save_hybrid=True' will append ground truth to predictions for autolabelling.\n"
+                "WARNING ⚠️ 'save_hybrid=True' will cause incorrect mAP.\n"
+            )
+
+    def preprocess(self, batch):
+        """Preprocesses batch of images for YOLO training."""
+        batch["img"] = batch["img"].to(self.device, non_blocking=True)
+        batch["img"] = (batch["img"].half() if self.args.half else batch["img"].float()) / 255
+        for k in ["batch_idx", "cls", "bboxes"]:
+            batch[k] = batch[k].to(self.device)
+
+        if self.args.save_hybrid:
+            height, width = batch["img"].shape[2:]
+            nb = len(batch["img"])
+            bboxes = batch["bboxes"] * torch.tensor((width, height, width, height), device=self.device)
+            self.lb = [
+                torch.cat([batch["cls"][batch["batch_idx"] == i], bboxes[batch["batch_idx"] == i]], dim=-1)
+                for i in range(nb)
+            ]
+
+        return batch
+
+    def init_metrics(self, model):
+        """Initialize evaluation metrics for YOLO."""
+        val = self.data.get(self.args.split, "")  # validation path
+        self.is_coco = (
+            isinstance(val, str)
+            and "coco" in val
+            and (val.endswith(f"{os.sep}val2017.txt") or val.endswith(f"{os.sep}test-dev2017.txt"))
+        )  # is COCO
+        self.is_lvis = isinstance(val, str) and "lvis" in val and not self.is_coco  # is LVIS
+        self.class_map = converter.coco80_to_coco91_class() if self.is_coco else list(range(1, len(model.names) + 1))
+        self.args.save_json |= self.args.val and (self.is_coco or self.is_lvis) and not self.training  # run final val
+        self.names = model.names
+        self.nc = len(model.names)
+        self.metrics.names = self.names
+        self.metrics.plot = self.args.plots
+        self.confusion_matrix = ConfusionMatrix(nc=self.nc, conf=self.args.conf)
+        self.seen = 0
+        self.jdict = []
+        self.stats = dict(tp=[], conf=[], pred_cls=[], target_cls=[], target_img=[])
+
+    def get_desc(self):
+        """Return a formatted string summarizing class metrics of YOLO model."""
+        return ("%22s" + "%11s" * 7) % ("Class", "Images", "Instances", "Box(P", "R", "mAP50", "mAP75","mAP50-95)")
+
+    def postprocess(self, preds):
+        """Apply Non-maximum suppression to prediction outputs."""
+        return ops.non_max_suppression(
+            preds,
+            self.args.conf,
+            self.args.iou,
+            labels=self.lb,
+            multi_label=True,
+            agnostic=self.args.single_cls or self.args.agnostic_nms,
+            max_det=self.args.max_det,
+        )
+
+    def _prepare_batch(self, si, batch):
+        """Prepares a batch of images and annotations for validation."""
+        idx = batch["batch_idx"] == si
+        cls = batch["cls"][idx].squeeze(-1)
+        bbox = batch["bboxes"][idx]
+        ori_shape = batch["ori_shape"][si]
+        imgsz = batch["img"].shape[2:]
+        ratio_pad = batch["ratio_pad"][si]
+        if len(cls):
+            bbox = ops.xywh2xyxy(bbox) * torch.tensor(imgsz, device=self.device)[[1, 0, 1, 0]]  # target boxes
+            ops.scale_boxes(imgsz, bbox, ori_shape, ratio_pad=ratio_pad)  # native-space labels
+        return {"cls": cls, "bbox": bbox, "ori_shape": ori_shape, "imgsz": imgsz, "ratio_pad": ratio_pad}
+
+    def _prepare_pred(self, pred, pbatch):
+        """Prepares a batch of images and annotations for validation."""
+        predn = pred.clone()
+        ops.scale_boxes(
+            pbatch["imgsz"], predn[:, :4], pbatch["ori_shape"], ratio_pad=pbatch["ratio_pad"]
+        )  # native-space pred
+        return predn
+
+    def update_metrics(self, preds, batch):
+        """Metrics."""
+        for si, pred in enumerate(preds):
+            self.seen += 1
+            npr = len(pred)
+            stat = dict(
+                conf=torch.zeros(0, device=self.device),
+                pred_cls=torch.zeros(0, device=self.device),
+                tp=torch.zeros(npr, self.niou, dtype=torch.bool, device=self.device),
+            )
+            pbatch = self._prepare_batch(si, batch)
+            cls, bbox = pbatch.pop("cls"), pbatch.pop("bbox")
+            nl = len(cls)
+            stat["target_cls"] = cls
+            stat["target_img"] = cls.unique()
+            if npr == 0:
+                if nl:
+                    for k in self.stats.keys():
+                        self.stats[k].append(stat[k])
+                    if self.args.plots:
+                        self.confusion_matrix.process_batch(detections=None, gt_bboxes=bbox, gt_cls=cls)
+                continue
+
+            # Predictions
+            if self.args.single_cls:
+                pred[:, 5] = 0
+            predn = self._prepare_pred(pred, pbatch)
+            stat["conf"] = predn[:, 4]
+            stat["pred_cls"] = predn[:, 5]
+
+            # Evaluate
+            if nl:
+                stat["tp"] = self._process_batch(predn, bbox, cls)
+            if self.args.plots:
+                self.confusion_matrix.process_batch(predn, bbox, cls)
+            for k in self.stats.keys():
+                self.stats[k].append(stat[k])
+
+            # Save
+            if self.args.save_json:
+                self.pred_to_json(predn, batch["im_file"][si])
+            if self.args.save_txt:
+                self.save_one_txt(
+                    predn,
+                    self.args.save_conf,
+                    pbatch["ori_shape"],
+                    self.save_dir / "labels" / f"{Path(batch['im_file'][si]).stem}.txt",
+                )
+
+    def finalize_metrics(self, *args, **kwargs):
+        """Set final values for metrics speed and confusion matrix."""
+        self.metrics.speed = self.speed
+        self.metrics.confusion_matrix = self.confusion_matrix
+
+    def get_stats(self):
+        """Returns metrics statistics and results dictionary."""
+        stats = {k: torch.cat(v, 0).cpu().numpy() for k, v in self.stats.items()}  # to numpy
+        self.nt_per_class = np.bincount(stats["target_cls"].astype(int), minlength=self.nc)
+        self.nt_per_image = np.bincount(stats["target_img"].astype(int), minlength=self.nc)
+        stats.pop("target_img", None)
+        if len(stats) and stats["tp"].any():
+            self.metrics.process(**stats)
+        return self.metrics.results_dict
+
+    def print_results(self):
+        """Prints training/validation set metrics per class."""
+        pf = "%22s" + "%11i" * 2 + "%11.3g" * len(self.metrics.keys)  # print format
+        LOGGER.info(pf % ("all", self.seen, self.nt_per_class.sum(), *self.metrics.mean_results()))
+        if self.nt_per_class.sum() == 0:
+            LOGGER.warning(f"WARNING ⚠️ no labels found in {self.args.task} set, can not compute metrics without labels")
+
+        # Print results per class
+        if self.args.verbose and not self.training and self.nc > 1 and len(self.stats):
+            for i, c in enumerate(self.metrics.ap_class_index):
+                LOGGER.info(
+                    pf % (self.names[c], self.nt_per_image[c], self.nt_per_class[c], *self.metrics.class_result(i))
+                )
+
+        if self.args.plots:
+            for normalize in True, False:
+                self.confusion_matrix.plot(
+                    save_dir=self.save_dir, names=self.names.values(), normalize=normalize, on_plot=self.on_plot
+                )
+
+    def _process_batch(self, detections, gt_bboxes, gt_cls):
+        """
+        Return correct prediction matrix.
+
+        Args:
+            detections (torch.Tensor): Tensor of shape (N, 6) representing detections where each detection is
+                (x1, y1, x2, y2, conf, class).
+            gt_bboxes (torch.Tensor): Tensor of shape (M, 4) representing ground-truth bounding box coordinates. Each
+                bounding box is of the format: (x1, y1, x2, y2).
+            gt_cls (torch.Tensor): Tensor of shape (M,) representing target class indices.
+
+        Returns:
+            (torch.Tensor): Correct prediction matrix of shape (N, 10) for 10 IoU levels.
+
+        Note:
+            The function does not return any value directly usable for metrics calculation. Instead, it provides an
+            intermediate representation used for evaluating predictions against ground truth.
+        """
+        iou = box_iou(gt_bboxes, detections[:, :4])
+        return self.match_predictions(detections[:, 5], gt_cls, iou)
+
+    def build_dataset(self, img_path, mode="val", batch=None):
+        """
+        Build YOLO Dataset.
+
+        Args:
+            img_path (str): Path to the folder containing images.
+            mode (str): `train` mode or `val` mode, users are able to customize different augmentations for each mode.
+            batch (int, optional): Size of batches, this is for `rect`. Defaults to None.
+        """
+        return build_yolo_dataset(self.args, img_path, batch, self.data, mode=mode, stride=self.stride)
+
+    def get_dataloader(self, dataset_path, batch_size):
+        """Construct and return dataloader."""
+        dataset = self.build_dataset(dataset_path, batch=batch_size, mode="val")
+        return build_dataloader(dataset, batch_size, self.args.workers, shuffle=False, rank=-1)  # return dataloader
+
+    def plot_val_samples(self, batch, ni):
+        """Plot validation image samples."""
+        plot_images(
+            batch["img"],
+            batch["batch_idx"],
+            batch["cls"].squeeze(-1),
+            batch["bboxes"],
+            paths=batch["im_file"],
+            fname=self.save_dir / f"val_batch{ni}_labels.jpg",
+            names=self.names,
+            on_plot=self.on_plot,
+        )
+
+    def plot_predictions(self, batch, preds, ni):
+        """Plots predicted bounding boxes on input images and saves the result."""
+        plot_images(
+            batch["img"],
+            *output_to_target(preds, max_det=self.args.max_det),
+            paths=batch["im_file"],
+            fname=self.save_dir / f"val_batch{ni}_pred.jpg",
+            names=self.names,
+            on_plot=self.on_plot,
+        )  # pred
+
+    def save_one_txt(self, predn, save_conf, shape, file):
+        """Save YOLO detections to a txt file in normalized coordinates in a specific format."""
+        from ultralytics.engine.results import Results
+
+        Results(
+            np.zeros((shape[0], shape[1]), dtype=np.uint8),
+            path=None,
+            names=self.names,
+            boxes=predn[:, :6],
+        ).save_txt(file, save_conf=save_conf)
+
+    def pred_to_json(self, predn, filename):
+        """Serialize YOLO predictions to COCO json format."""
+        stem = Path(filename).stem
+        image_id = int(stem) if stem.isnumeric() else stem
+        box = ops.xyxy2xywh(predn[:, :4])  # xywh
+        box[:, :2] -= box[:, 2:] / 2  # xy center to top-left corner
+        for p, b in zip(predn.tolist(), box.tolist()):
+            self.jdict.append(
+                {
+                    "image_id": image_id,
+                    "category_id": self.class_map[int(p[5])],
+                    "bbox": [round(x, 3) for x in b],
+                    "score": round(p[4], 5),
+                }
+            )
+
+    def eval_json(self, stats):
+        """Evaluates YOLO output in JSON format and returns performance statistics."""
+        if self.args.save_json and (self.is_coco or self.is_lvis) and len(self.jdict):
+            pred_json = self.save_dir / "predictions.json"  # predictions
+            anno_json = (
+                self.data["path"]
+                / "annotations"
+                / ("instances_val2017.json" if self.is_coco else f"lvis_v1_{self.args.split}.json")
+            )  # annotations
+            pkg = "pycocotools" if self.is_coco else "lvis"
+            LOGGER.info(f"\nEvaluating {pkg} mAP using {pred_json} and {anno_json}...")
+            try:  # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb
+                for x in pred_json, anno_json:
+                    assert x.is_file(), f"{x} file not found"
+                check_requirements("pycocotools>=2.0.6" if self.is_coco else "lvis>=0.5.3")
+                if self.is_coco:
+                    from pycocotools.coco import COCO  # noqa
+                    from pycocotools.cocoeval import COCOeval  # noqa
+
+                    anno = COCO(str(anno_json))  # init annotations api
+                    pred = anno.loadRes(str(pred_json))  # init predictions api (must pass string, not Path)
+                    val = COCOeval(anno, pred, "bbox")
+                else:
+                    from lvis import LVIS, LVISEval
+
+                    anno = LVIS(str(anno_json))  # init annotations api
+                    pred = anno._load_json(str(pred_json))  # init predictions api (must pass string, not Path)
+                    val = LVISEval(anno, pred, "bbox")
+                val.params.imgIds = [int(Path(x).stem) for x in self.dataloader.dataset.im_files]  # images to eval
+                val.evaluate()
+                val.accumulate()
+                val.summarize()
+                if self.is_lvis:
+                    val.print_results()  # explicitly call print_results
+                # update mAP50-95 and mAP50
+                stats[self.metrics.keys[-1]], stats[self.metrics.keys[-2]] = (
+                    val.stats[:2] if self.is_coco else [val.results["AP50"], val.results["AP"]]
+                )
+            except Exception as e:
+                LOGGER.warning(f"{pkg} unable to run: {e}")
+        return stats

ultralytics/models/yolo/model.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from pathlib import Path
+
+from ultralytics.engine.model import Model
+from ultralytics.models import yolo
+from ultralytics.nn.tasks import ClassificationModel, DetectionModel, OBBModel, PoseModel, SegmentationModel, WorldModel
+from ultralytics.utils import ROOT, yaml_load
+
+
+class YOLO(Model):
+    """YOLO (You Only Look Once) object detection model."""
+
+    def __init__(self, model="yolo11n.pt", task=None, verbose=False):
+        """Initialize YOLO model, switching to YOLOWorld if model filename contains '-world'."""
+        path = Path(model)
+        if "-world" in path.stem and path.suffix in {".pt", ".yaml", ".yml"}:  # if YOLOWorld PyTorch model
+            new_instance = YOLOWorld(path, verbose=verbose)
+            self.__class__ = type(new_instance)
+            self.__dict__ = new_instance.__dict__
+        else:
+            # Continue with default YOLO initialization
+            super().__init__(model=model, task=task, verbose=verbose)
+
+    @property
+    def task_map(self):
+        """Map head to model, trainer, validator, and predictor classes."""
+        return {
+            "classify": {
+                "model": ClassificationModel,
+                "trainer": yolo.classify.ClassificationTrainer,
+                "validator": yolo.classify.ClassificationValidator,
+                "predictor": yolo.classify.ClassificationPredictor,
+            },
+            "detect": {
+                "model": DetectionModel,
+                "trainer": yolo.detect.DetectionTrainer,
+                "validator": yolo.detect.DetectionValidator,
+                "predictor": yolo.detect.DetectionPredictor,
+            },
+            "segment": {
+                "model": SegmentationModel,
+                "trainer": yolo.segment.SegmentationTrainer,
+                "validator": yolo.segment.SegmentationValidator,
+                "predictor": yolo.segment.SegmentationPredictor,
+            },
+            "pose": {
+                "model": PoseModel,
+                "trainer": yolo.pose.PoseTrainer,
+                "validator": yolo.pose.PoseValidator,
+                "predictor": yolo.pose.PosePredictor,
+            },
+            "obb": {
+                "model": OBBModel,
+                "trainer": yolo.obb.OBBTrainer,
+                "validator": yolo.obb.OBBValidator,
+                "predictor": yolo.obb.OBBPredictor,
+            },
+        }
+
+
+class YOLOWorld(Model):
+    """YOLO-World object detection model."""
+
+    def __init__(self, model="yolov8s-world.pt", verbose=False) -> None:
+        """
+        Initialize YOLOv8-World model with a pre-trained model file.
+
+        Loads a YOLOv8-World model for object detection. If no custom class names are provided, it assigns default
+        COCO class names.
+
+        Args:
+            model (str | Path): Path to the pre-trained model file. Supports *.pt and *.yaml formats.
+            verbose (bool): If True, prints additional information during initialization.
+        """
+        super().__init__(model=model, task="detect", verbose=verbose)
+
+        # Assign default COCO class names when there are no custom names
+        if not hasattr(self.model, "names"):
+            self.model.names = yaml_load(ROOT / "cfg/datasets/coco8.yaml").get("names")
+
+    @property
+    def task_map(self):
+        """Map head to model, validator, and predictor classes."""
+        return {
+            "detect": {
+                "model": WorldModel,
+                "validator": yolo.detect.DetectionValidator,
+                "predictor": yolo.detect.DetectionPredictor,
+                "trainer": yolo.world.WorldTrainer,
+            }
+        }
+
+    def set_classes(self, classes):
+        """
+        Set classes.
+
+        Args:
+            classes (List(str)): A list of categories i.e. ["person"].
+        """
+        self.model.set_classes(classes)
+        # Remove background if it's given
+        background = " "
+        if background in classes:
+            classes.remove(background)
+        self.model.names = classes
+
+        # Reset method class names
+        # self.predictor = None  # reset predictor otherwise old names remain
+        if self.predictor:
+            self.predictor.model.names = classes

ultralytics/models/yolo/obb/__init__.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from .predict import OBBPredictor
+from .train import OBBTrainer
+from .val import OBBValidator
+
+__all__ = "OBBPredictor", "OBBTrainer", "OBBValidator"

ultralytics/models/yolo/obb/predict.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+import torch
+
+from ultralytics.engine.results import Results
+from ultralytics.models.yolo.detect.predict import DetectionPredictor
+from ultralytics.utils import DEFAULT_CFG, ops
+
+
+class OBBPredictor(DetectionPredictor):
+    """
+    A class extending the DetectionPredictor class for prediction based on an Oriented Bounding Box (OBB) model.
+
+    Example:
+        ```python
+        from ultralytics.utils import ASSETS
+        from ultralytics.models.yolo.obb import OBBPredictor
+
+        args = dict(model="yolov8n-obb.pt", source=ASSETS)
+        predictor = OBBPredictor(overrides=args)
+        predictor.predict_cli()
+        ```
+    """
+
+    def __init__(self, cfg=DEFAULT_CFG, overrides=None, _callbacks=None):
+        """Initializes OBBPredictor with optional model and data configuration overrides."""
+        super().__init__(cfg, overrides, _callbacks)
+        self.args.task = "obb"
+
+    def postprocess(self, preds, img, orig_imgs):
+        """Post-processes predictions and returns a list of Results objects."""
+        preds = ops.non_max_suppression(
+            preds,
+            self.args.conf,
+            self.args.iou,
+            agnostic=self.args.agnostic_nms,
+            max_det=self.args.max_det,
+            nc=len(self.model.names),
+            classes=self.args.classes,
+            rotated=True,
+        )
+
+        if not isinstance(orig_imgs, list):  # input images are a torch.Tensor, not a list
+            orig_imgs = ops.convert_torch2numpy_batch(orig_imgs)
+
+        results = []
+        for pred, orig_img, img_path in zip(preds, orig_imgs, self.batch[0]):
+            rboxes = ops.regularize_rboxes(torch.cat([pred[:, :4], pred[:, -1:]], dim=-1))
+            rboxes[:, :4] = ops.scale_boxes(img.shape[2:], rboxes[:, :4], orig_img.shape, xywh=True)
+            # xywh, r, conf, cls
+            obb = torch.cat([rboxes, pred[:, 4:6]], dim=-1)
+            results.append(Results(orig_img, path=img_path, names=self.model.names, obb=obb))
+        return results

ultralytics/models/yolo/obb/train.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from copy import copy
+
+from ultralytics.models import yolo
+from ultralytics.nn.tasks import OBBModel
+from ultralytics.utils import DEFAULT_CFG, RANK
+
+
+class OBBTrainer(yolo.detect.DetectionTrainer):
+    """
+    A class extending the DetectionTrainer class for training based on an Oriented Bounding Box (OBB) model.
+
+    Example:
+        ```python
+        from ultralytics.models.yolo.obb import OBBTrainer
+
+        args = dict(model="yolov8n-obb.pt", data="dota8.yaml", epochs=3)
+        trainer = OBBTrainer(overrides=args)
+        trainer.train()
+        ```
+    """
+
+    def __init__(self, cfg=DEFAULT_CFG, overrides=None, _callbacks=None):
+        """Initialize a OBBTrainer object with given arguments."""
+        if overrides is None:
+            overrides = {}
+        overrides["task"] = "obb"
+        super().__init__(cfg, overrides, _callbacks)
+
+    def get_model(self, cfg=None, weights=None, verbose=True):
+        """Return OBBModel initialized with specified config and weights."""
+        model = OBBModel(cfg, ch=3, nc=self.data["nc"], verbose=verbose and RANK == -1)
+        if weights:
+            model.load(weights)
+
+        return model
+
+    def get_validator(self):
+        """Return an instance of OBBValidator for validation of YOLO model."""
+        self.loss_names = "box_loss", "cls_loss", "dfl_loss"
+        return yolo.obb.OBBValidator(
+            self.test_loader, save_dir=self.save_dir, args=copy(self.args), _callbacks=self.callbacks
+        )

ultralytics/models/yolo/obb/val.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from pathlib import Path
+
+import torch
+
+from ultralytics.models.yolo.detect import DetectionValidator
+from ultralytics.utils import LOGGER, ops
+from ultralytics.utils.metrics import OBBMetrics, batch_probiou
+from ultralytics.utils.plotting import output_to_rotated_target, plot_images
+
+
+class OBBValidator(DetectionValidator):
+    """
+    A class extending the DetectionValidator class for validation based on an Oriented Bounding Box (OBB) model.
+
+    Example:
+        ```python
+        from ultralytics.models.yolo.obb import OBBValidator
+
+        args = dict(model="yolov8n-obb.pt", data="dota8.yaml")
+        validator = OBBValidator(args=args)
+        validator(model=args["model"])
+        ```
+    """
+
+    def __init__(self, dataloader=None, save_dir=None, pbar=None, args=None, _callbacks=None):
+        """Initialize OBBValidator and set task to 'obb', metrics to OBBMetrics."""
+        super().__init__(dataloader, save_dir, pbar, args, _callbacks)
+        self.args.task = "obb"
+        self.metrics = OBBMetrics(save_dir=self.save_dir, plot=True, on_plot=self.on_plot)
+
+    def init_metrics(self, model):
+        """Initialize evaluation metrics for YOLO."""
+        super().init_metrics(model)
+        val = self.data.get(self.args.split, "")  # validation path
+        self.is_dota = isinstance(val, str) and "DOTA" in val  # is COCO
+
+    def postprocess(self, preds):
+        """Apply Non-maximum suppression to prediction outputs."""
+        return ops.non_max_suppression(
+            preds,
+            self.args.conf,
+            self.args.iou,
+            labels=self.lb,
+            nc=self.nc,
+            multi_label=True,
+            agnostic=self.args.single_cls or self.args.agnostic_nms,
+            max_det=self.args.max_det,
+            rotated=True,
+        )
+
+    def _process_batch(self, detections, gt_bboxes, gt_cls):
+        """
+        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.
+
+        Example:
+            ```python
+            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.
+        """
+        iou = batch_probiou(gt_bboxes, torch.cat([detections[:, :4], detections[:, -1:]], dim=-1))
+        return self.match_predictions(detections[:, 5], gt_cls, iou)
+
+    def _prepare_batch(self, si, batch):
+        """Prepares and returns a batch for OBB validation."""
+        idx = batch["batch_idx"] == si
+        cls = batch["cls"][idx].squeeze(-1)
+        bbox = batch["bboxes"][idx]
+        ori_shape = batch["ori_shape"][si]
+        imgsz = batch["img"].shape[2:]
+        ratio_pad = batch["ratio_pad"][si]
+        if len(cls):
+            bbox[..., :4].mul_(torch.tensor(imgsz, device=self.device)[[1, 0, 1, 0]])  # target boxes
+            ops.scale_boxes(imgsz, bbox, ori_shape, ratio_pad=ratio_pad, xywh=True)  # native-space labels
+        return {"cls": cls, "bbox": bbox, "ori_shape": ori_shape, "imgsz": imgsz, "ratio_pad": ratio_pad}
+
+    def _prepare_pred(self, pred, pbatch):
+        """Prepares and returns a batch for OBB validation with scaled and padded bounding boxes."""
+        predn = pred.clone()
+        ops.scale_boxes(
+            pbatch["imgsz"], predn[:, :4], pbatch["ori_shape"], ratio_pad=pbatch["ratio_pad"], xywh=True
+        )  # native-space pred
+        return predn
+
+    def plot_predictions(self, batch, preds, ni):
+        """Plots predicted bounding boxes on input images and saves the result."""
+        plot_images(
+            batch["img"],
+            *output_to_rotated_target(preds, max_det=self.args.max_det),
+            paths=batch["im_file"],
+            fname=self.save_dir / f"val_batch{ni}_pred.jpg",
+            names=self.names,
+            on_plot=self.on_plot,
+        )  # pred
+
+    def pred_to_json(self, predn, filename):
+        """Serialize YOLO predictions to COCO json format."""
+        stem = Path(filename).stem
+        image_id = int(stem) if stem.isnumeric() else stem
+        rbox = torch.cat([predn[:, :4], predn[:, -1:]], dim=-1)
+        poly = ops.xywhr2xyxyxyxy(rbox).view(-1, 8)
+        for i, (r, b) in enumerate(zip(rbox.tolist(), poly.tolist())):
+            self.jdict.append(
+                {
+                    "image_id": image_id,
+                    "category_id": self.class_map[int(predn[i, 5].item())],
+                    "score": round(predn[i, 4].item(), 5),
+                    "rbox": [round(x, 3) for x in r],
+                    "poly": [round(x, 3) for x in b],
+                }
+            )
+
+    def save_one_txt(self, predn, save_conf, shape, file):
+        """Save YOLO detections to a txt file in normalized coordinates in a specific format."""
+        import numpy as np
+
+        from ultralytics.engine.results import Results
+
+        rboxes = torch.cat([predn[:, :4], predn[:, -1:]], dim=-1)
+        # xywh, r, conf, cls
+        obb = torch.cat([rboxes, predn[:, 4:6]], dim=-1)
+        Results(
+            np.zeros((shape[0], shape[1]), dtype=np.uint8),
+            path=None,
+            names=self.names,
+            obb=obb,
+        ).save_txt(file, save_conf=save_conf)
+
+    def eval_json(self, stats):
+        """Evaluates YOLO output in JSON format and returns performance statistics."""
+        if self.args.save_json and self.is_dota and len(self.jdict):
+            import json
+            import re
+            from collections import defaultdict
+
+            pred_json = self.save_dir / "predictions.json"  # predictions
+            pred_txt = self.save_dir / "predictions_txt"  # predictions
+            pred_txt.mkdir(parents=True, exist_ok=True)
+            data = json.load(open(pred_json))
+            # Save split results
+            LOGGER.info(f"Saving predictions with DOTA format to {pred_txt}...")
+            for d in data:
+                image_id = d["image_id"]
+                score = d["score"]
+                classname = self.names[d["category_id"] - 1].replace(" ", "-")
+                p = d["poly"]
+
+                with open(f"{pred_txt / f'Task1_{classname}'}.txt", "a") as f:
+                    f.writelines(f"{image_id} {score} {p[0]} {p[1]} {p[2]} {p[3]} {p[4]} {p[5]} {p[6]} {p[7]}\n")
+            # Save merged results, this could result slightly lower map than using official merging script,
+            # because of the probiou calculation.
+            pred_merged_txt = self.save_dir / "predictions_merged_txt"  # predictions
+            pred_merged_txt.mkdir(parents=True, exist_ok=True)
+            merged_results = defaultdict(list)
+            LOGGER.info(f"Saving merged predictions with DOTA format to {pred_merged_txt}...")
+            for d in data:
+                image_id = d["image_id"].split("__")[0]
+                pattern = re.compile(r"\d+___\d+")
+                x, y = (int(c) for c in re.findall(pattern, d["image_id"])[0].split("___"))
+                bbox, score, cls = d["rbox"], d["score"], d["category_id"] - 1
+                bbox[0] += x
+                bbox[1] += y
+                bbox.extend([score, cls])
+                merged_results[image_id].append(bbox)
+            for image_id, bbox in merged_results.items():
+                bbox = torch.tensor(bbox)
+                max_wh = torch.max(bbox[:, :2]).item() * 2
+                c = bbox[:, 6:7] * max_wh  # classes
+                scores = bbox[:, 5]  # scores
+                b = bbox[:, :5].clone()
+                b[:, :2] += c
+                # 0.3 could get results close to the ones from official merging script, even slightly better.
+                i = ops.nms_rotated(b, scores, 0.3)
+                bbox = bbox[i]
+
+                b = ops.xywhr2xyxyxyxy(bbox[:, :5]).view(-1, 8)
+                for x in torch.cat([b, bbox[:, 5:7]], dim=-1).tolist():
+                    classname = self.names[int(x[-1])].replace(" ", "-")
+                    p = [round(i, 3) for i in x[:-2]]  # poly
+                    score = round(x[-2], 3)
+
+                    with open(f"{pred_merged_txt / f'Task1_{classname}'}.txt", "a") as f:
+                        f.writelines(f"{image_id} {score} {p[0]} {p[1]} {p[2]} {p[3]} {p[4]} {p[5]} {p[6]} {p[7]}\n")
+
+        return stats

ultralytics/models/yolo/pose/__init__.py

+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from .predict import PosePredictor
+from .train import PoseTrainer
+from .val import PoseValidator
+
+__all__ = "PoseTrainer", "PoseValidator", "PosePredictor"