Tài liệu tham khảo cho ultralytics/engine/results.py

class BaseTensor(SimpleClass):
    """Base tensor class with additional methods for easy manipulation and device handling."""

    def __init__(self, data, orig_shape) -> None:
        """
        Initialize BaseTensor with data and original shape.

        Args:
            data (torch.Tensor | np.ndarray): Predictions, such as bboxes, masks and keypoints.
            orig_shape (tuple): Original shape of image.
        """
        assert isinstance(data, (torch.Tensor, np.ndarray))
        self.data = data
        self.orig_shape = orig_shape

    @property
    def shape(self):
        """Return the shape of the data tensor."""
        return self.data.shape

    def cpu(self):
        """Return a copy of the tensor on CPU memory."""
        return self if isinstance(self.data, np.ndarray) else self.__class__(self.data.cpu(), self.orig_shape)

    def numpy(self):
        """Return a copy of the tensor as a numpy array."""
        return self if isinstance(self.data, np.ndarray) else self.__class__(self.data.numpy(), self.orig_shape)

    def cuda(self):
        """Return a copy of the tensor on GPU memory."""
        return self.__class__(torch.as_tensor(self.data).cuda(), self.orig_shape)

    def to(self, *args, **kwargs):
        """Return a copy of the tensor with the specified device and dtype."""
        return self.__class__(torch.as_tensor(self.data).to(*args, **kwargs), self.orig_shape)

    def __len__(self):  # override len(results)
        """Return the length of the data tensor."""
        return len(self.data)

    def __getitem__(self, idx):
        """Return a BaseTensor with the specified index of the data tensor."""
        return self.__class__(self.data[idx], self.orig_shape)

class Results(SimpleClass):
    """
    A class for storing and manipulating inference results.

    Attributes:
        orig_img (numpy.ndarray): Original image as a numpy array.
        orig_shape (tuple): Original image shape in (height, width) format.
        boxes (Boxes, optional): Object containing detection bounding boxes.
        masks (Masks, optional): Object containing detection masks.
        probs (Probs, optional): Object containing class probabilities for classification tasks.
        keypoints (Keypoints, optional): Object containing detected keypoints for each object.
        speed (dict): Dictionary of preprocess, inference, and postprocess speeds (ms/image).
        names (dict): Dictionary of class names.
        path (str): Path to the image file.

    Methods:
        update(boxes=None, masks=None, probs=None, obb=None): Updates object attributes with new detection results.
        cpu(): Returns a copy of the Results object with all tensors on CPU memory.
        numpy(): Returns a copy of the Results object with all tensors as numpy arrays.
        cuda(): Returns a copy of the Results object with all tensors on GPU memory.
        to(*args, **kwargs): Returns a copy of the Results object with tensors on a specified device and dtype.
        new(): Returns a new Results object with the same image, path, and names.
        plot(...): Plots detection results on an input image, returning an annotated image.
        show(): Show annotated results to screen.
        save(filename): Save annotated results to file.
        verbose(): Returns a log string for each task, detailing detections and classifications.
        save_txt(txt_file, save_conf=False): Saves detection results to a text file.
        save_crop(save_dir, file_name=Path("im.jpg")): Saves cropped detection images.
        tojson(normalize=False): Converts detection results to JSON format.
    """

    def __init__(self, orig_img, path, names, boxes=None, masks=None, probs=None, keypoints=None, obb=None) -> None:
        """
        Initialize the Results class.

        Args:
            orig_img (numpy.ndarray): The original image as a numpy array.
            path (str): The path to the image file.
            names (dict): A dictionary of class names.
            boxes (torch.tensor, optional): A 2D tensor of bounding box coordinates for each detection.
            masks (torch.tensor, optional): A 3D tensor of detection masks, where each mask is a binary image.
            probs (torch.tensor, optional): A 1D tensor of probabilities of each class for classification task.
            keypoints (torch.tensor, optional): A 2D tensor of keypoint coordinates for each detection.
            obb (torch.tensor, optional): A 2D tensor of oriented bounding box coordinates for each detection.
        """
        self.orig_img = orig_img
        self.orig_shape = orig_img.shape[:2]
        self.boxes = Boxes(boxes, self.orig_shape) if boxes is not None else None  # native size boxes
        self.masks = Masks(masks, self.orig_shape) if masks is not None else None  # native size or imgsz masks
        self.probs = Probs(probs) if probs is not None else None
        self.keypoints = Keypoints(keypoints, self.orig_shape) if keypoints is not None else None
        self.obb = OBB(obb, self.orig_shape) if obb is not None else None
        self.speed = {"preprocess": None, "inference": None, "postprocess": None}  # milliseconds per image
        self.names = names
        self.path = path
        self.save_dir = None
        self._keys = "boxes", "masks", "probs", "keypoints", "obb"

    def __getitem__(self, idx):
        """Return a Results object for the specified index."""
        return self._apply("__getitem__", idx)

    def __len__(self):
        """Return the number of detections in the Results object."""
        for k in self._keys:
            v = getattr(self, k)
            if v is not None:
                return len(v)

    def update(self, boxes=None, masks=None, probs=None, obb=None):
        """Update the boxes, masks, and probs attributes of the Results object."""
        if boxes is not None:
            self.boxes = Boxes(ops.clip_boxes(boxes, self.orig_shape), self.orig_shape)
        if masks is not None:
            self.masks = Masks(masks, self.orig_shape)
        if probs is not None:
            self.probs = probs
        if obb is not None:
            self.obb = OBB(obb, self.orig_shape)

    def _apply(self, fn, *args, **kwargs):
        """
        Applies a function to all non-empty attributes and returns a new Results object with modified attributes. This
        function is internally called by methods like .to(), .cuda(), .cpu(), etc.

        Args:
            fn (str): The name of the function to apply.
            *args: Variable length argument list to pass to the function.
            **kwargs: Arbitrary keyword arguments to pass to the function.

        Returns:
            Results: A new Results object with attributes modified by the applied function.
        """
        r = self.new()
        for k in self._keys:
            v = getattr(self, k)
            if v is not None:
                setattr(r, k, getattr(v, fn)(*args, **kwargs))
        return r

    def cpu(self):
        """Return a copy of the Results object with all tensors on CPU memory."""
        return self._apply("cpu")

    def numpy(self):
        """Return a copy of the Results object with all tensors as numpy arrays."""
        return self._apply("numpy")

    def cuda(self):
        """Return a copy of the Results object with all tensors on GPU memory."""
        return self._apply("cuda")

    def to(self, *args, **kwargs):
        """Return a copy of the Results object with tensors on the specified device and dtype."""
        return self._apply("to", *args, **kwargs)

    def new(self):
        """Return a new Results object with the same image, path, and names."""
        return Results(orig_img=self.orig_img, path=self.path, names=self.names)

    def plot(
        self,
        conf=True,
        line_width=None,
        font_size=None,
        font="Arial.ttf",
        pil=False,
        img=None,
        im_gpu=None,
        kpt_radius=5,
        kpt_line=True,
        labels=True,
        boxes=True,
        masks=True,
        probs=True,
        show=False,
        save=False,
        filename=None,
    ):
        """
        Plots the detection results on an input RGB image. Accepts a numpy array (cv2) or a PIL Image.

        Args:
            conf (bool): Whether to plot the detection confidence score.
            line_width (float, optional): The line width of the bounding boxes. If None, it is scaled to the image size.
            font_size (float, optional): The font size of the text. If None, it is scaled to the image size.
            font (str): The font to use for the text.
            pil (bool): Whether to return the image as a PIL Image.
            img (numpy.ndarray): Plot to another image. if not, plot to original image.
            im_gpu (torch.Tensor): Normalized image in gpu with shape (1, 3, 640, 640), for faster mask plotting.
            kpt_radius (int, optional): Radius of the drawn keypoints. Default is 5.
            kpt_line (bool): Whether to draw lines connecting keypoints.
            labels (bool): Whether to plot the label of bounding boxes.
            boxes (bool): Whether to plot the bounding boxes.
            masks (bool): Whether to plot the masks.
            probs (bool): Whether to plot classification probability
            show (bool): Whether to display the annotated image directly.
            save (bool): Whether to save the annotated image to `filename`.
            filename (str): Filename to save image to if save is True.

        Returns:
            (numpy.ndarray): A numpy array of the annotated image.

        Example:
            ```python
            from PIL import Image
            from ultralytics import YOLO

            model = YOLO('yolov8n.pt')
            results = model('bus.jpg')  # results list
            for r in results:
                im_array = r.plot()  # plot a BGR numpy array of predictions
                im = Image.fromarray(im_array[..., ::-1])  # RGB PIL image
                im.show()  # show image
                im.save('results.jpg')  # save image
            ```
        """
        if img is None and isinstance(self.orig_img, torch.Tensor):
            img = (self.orig_img[0].detach().permute(1, 2, 0).contiguous() * 255).to(torch.uint8).cpu().numpy()

        names = self.names
        is_obb = self.obb is not None
        pred_boxes, show_boxes = self.obb if is_obb else self.boxes, boxes
        pred_masks, show_masks = self.masks, masks
        pred_probs, show_probs = self.probs, probs
        annotator = Annotator(
            deepcopy(self.orig_img if img is None else img),
            line_width,
            font_size,
            font,
            pil or (pred_probs is not None and show_probs),  # Classify tasks default to pil=True
            example=names,
        )

        # Plot Segment results
        if pred_masks and show_masks:
            if im_gpu is None:
                img = LetterBox(pred_masks.shape[1:])(image=annotator.result())
                im_gpu = (
                    torch.as_tensor(img, dtype=torch.float16, device=pred_masks.data.device)
                    .permute(2, 0, 1)
                    .flip(0)
                    .contiguous()
                    / 255
                )
            idx = pred_boxes.cls if pred_boxes else range(len(pred_masks))
            annotator.masks(pred_masks.data, colors=[colors(x, True) for x in idx], im_gpu=im_gpu)

        # Plot Detect results
        if pred_boxes is not None and show_boxes:
            for d in reversed(pred_boxes):
                c, conf, id = int(d.cls), float(d.conf) if conf else None, None if d.id is None else int(d.id.item())
                name = ("" if id is None else f"id:{id} ") + names[c]
                label = (f"{name} {conf:.2f}" if conf else name) if labels else None
                box = d.xyxyxyxy.reshape(-1, 4, 2).squeeze() if is_obb else d.xyxy.squeeze()
                annotator.box_label(box, label, color=colors(c, True), rotated=is_obb)

        # Plot Classify results
        if pred_probs is not None and show_probs:
            text = ",\n".join(f"{names[j] if names else j} {pred_probs.data[j]:.2f}" for j in pred_probs.top5)
            x = round(self.orig_shape[0] * 0.03)
            annotator.text([x, x], text, txt_color=(255, 255, 255))  # TODO: allow setting colors

        # Plot Pose results
        if self.keypoints is not None:
            for k in reversed(self.keypoints.data):
                annotator.kpts(k, self.orig_shape, radius=kpt_radius, kpt_line=kpt_line)

        # Show results
        if show:
            annotator.show(self.path)

        # Save results
        if save:
            annotator.save(filename)

        return annotator.result()

    def show(self, *args, **kwargs):
        """Show annotated results image."""
        self.plot(show=True, *args, **kwargs)

    def save(self, filename=None, *args, **kwargs):
        """Save annotated results image."""
        if not filename:
            filename = f"results_{Path(self.path).name}"
        self.plot(save=True, filename=filename, *args, **kwargs)
        return filename

    def verbose(self):
        """Return log string for each task."""
        log_string = ""
        probs = self.probs
        boxes = self.boxes
        if len(self) == 0:
            return log_string if probs is not None else f"{log_string}(no detections), "
        if probs is not None:
            log_string += f"{', '.join(f'{self.names[j]} {probs.data[j]:.2f}' for j in probs.top5)}, "
        if boxes:
            for c in boxes.cls.unique():
                n = (boxes.cls == c).sum()  # detections per class
                log_string += f"{n} {self.names[int(c)]}{'s' * (n > 1)}, "
        return log_string

    def save_txt(self, txt_file, save_conf=False):
        """
        Save predictions into txt file.

        Args:
            txt_file (str): txt file path.
            save_conf (bool): save confidence score or not.
        """
        is_obb = self.obb is not None
        boxes = self.obb if is_obb else self.boxes
        masks = self.masks
        probs = self.probs
        kpts = self.keypoints
        texts = []
        if probs is not None:
            # Classify
            [texts.append(f"{probs.data[j]:.2f} {self.names[j]}") for j in probs.top5]
        elif boxes:
            # Detect/segment/pose
            for j, d in enumerate(boxes):
                c, conf, id = int(d.cls), float(d.conf), None if d.id is None else int(d.id.item())
                line = (c, *(d.xyxyxyxyn.view(-1) if is_obb else d.xywhn.view(-1)))
                if masks:
                    seg = masks[j].xyn[0].copy().reshape(-1)  # reversed mask.xyn, (n,2) to (n*2)
                    line = (c, *seg)
                if kpts is not None:
                    kpt = torch.cat((kpts[j].xyn, kpts[j].conf[..., None]), 2) if kpts[j].has_visible else kpts[j].xyn
                    line += (*kpt.reshape(-1).tolist(),)
                line += (conf,) * save_conf + (() if id is None else (id,))
                texts.append(("%g " * len(line)).rstrip() % line)

        if texts:
            Path(txt_file).parent.mkdir(parents=True, exist_ok=True)  # make directory
            with open(txt_file, "a") as f:
                f.writelines(text + "\n" for text in texts)

    def save_crop(self, save_dir, file_name=Path("im.jpg")):
        """
        Save cropped predictions to `save_dir/cls/file_name.jpg`.

        Args:
            save_dir (str | pathlib.Path): Save path.
            file_name (str | pathlib.Path): File name.
        """
        if self.probs is not None:
            LOGGER.warning("WARNING ⚠️ Classify task do not support `save_crop`.")
            return
        if self.obb is not None:
            LOGGER.warning("WARNING ⚠️ OBB task do not support `save_crop`.")
            return
        for d in self.boxes:
            save_one_box(
                d.xyxy,
                self.orig_img.copy(),
                file=Path(save_dir) / self.names[int(d.cls)] / f"{Path(file_name)}.jpg",
                BGR=True,
            )

    def summary(self, normalize=False, decimals=5):
        """Convert the results to a summarized format."""
        # Create list of detection dictionaries
        results = []
        if self.probs is not None:
            class_id = self.probs.top1
            results.append(
                {
                    "name": self.names[class_id],
                    "class": class_id,
                    "confidence": round(self.probs.top1conf.item(), decimals),
                }
            )
            return results

        data = self.boxes or self.obb
        is_obb = self.obb is not None
        h, w = self.orig_shape if normalize else (1, 1)
        for i, row in enumerate(data):  # xyxy, track_id if tracking, conf, class_id
            class_id, conf = int(row.cls), round(row.conf.item(), decimals)
            box = (row.xyxyxyxy if is_obb else row.xyxy).squeeze().reshape(-1, 2).tolist()
            xy = {}
            for j, b in enumerate(box):
                xy[f"x{j + 1}"] = round(b[0] / w, decimals)
                xy[f"y{j + 1}"] = round(b[1] / h, decimals)
            result = {"name": self.names[class_id], "class": class_id, "confidence": conf, "box": xy}
            if data.is_track:
                result["track_id"] = int(row.id.item())  # track ID
            if self.masks:
                result["segments"] = {
                    "x": (self.masks.xy[i][:, 0] / w).round(decimals).tolist(),
                    "y": (self.masks.xy[i][:, 1] / h).round(decimals).tolist(),
                }
            if self.keypoints is not None:
                x, y, visible = self.keypoints[i].data[0].cpu().unbind(dim=1)  # torch Tensor
                result["keypoints"] = {
                    "x": (x / w).numpy().round(decimals).tolist(),  # decimals named argument required
                    "y": (y / h).numpy().round(decimals).tolist(),
                    "visible": visible.numpy().round(decimals).tolist(),
                }
            results.append(result)

        return results

    def tojson(self, normalize=False, decimals=5):
        """Convert the results to JSON format."""
        import json

        return json.dumps(self.summary(normalize=normalize, decimals=decimals), indent=2)