dgenerate-ultralytics-headless 8.3.134__py3-none-any.whl

ultralytics/nn/text_model.py

@@ -0,0 +1,351 @@
+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from abc import abstractmethod
+from pathlib import Path
+
+import torch
+import torch.nn as nn
+
+from ultralytics.utils import checks
+from ultralytics.utils.torch_utils import smart_inference_mode
+
+try:
+    import clip
+except ImportError:
+    checks.check_requirements("git+https://github.com/ultralytics/CLIP.git")
+    import clip
+
+
+class TextModel(nn.Module):
+    """
+    Abstract base class for text encoding models.
+
+    This class defines the interface for text encoding models used in vision-language tasks. Subclasses must implement
+    the tokenize and encode_text methods.
+
+    Methods:
+        tokenize: Convert input texts to tokens.
+        encode_text: Encode tokenized texts into feature vectors.
+    """
+
+    def __init__(self):
+        """Initialize the TextModel base class."""
+        super().__init__()
+
+    @abstractmethod
+    def tokenize(texts):
+        """Convert input texts to tokens for model processing."""
+        pass
+
+    @abstractmethod
+    def encode_text(texts, dtype):
+        """Encode tokenized texts into normalized feature vectors."""
+        pass
+
+
+class CLIP(TextModel):
+    """
+    Implements OpenAI's CLIP (Contrastive Language-Image Pre-training) text encoder.
+
+    This class provides a text encoder based on OpenAI's CLIP model, which can convert text into feature vectors
+    that are aligned with corresponding image features in a shared embedding space.
+
+    Attributes:
+        model (clip.model.CLIP): The loaded CLIP model.
+        device (torch.device): Device where the model is loaded.
+
+    Methods:
+        tokenize: Convert input texts to CLIP tokens.
+        encode_text: Encode tokenized texts into normalized feature vectors.
+
+    Examples:
+        >>> from ultralytics.models.sam import CLIP
+        >>> import torch
+        >>> device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        >>> clip_model = CLIP(size="ViT-B/32", device=device)
+        >>> tokens = clip_model.tokenize(["a photo of a cat", "a photo of a dog"])
+        >>> text_features = clip_model.encode_text(tokens)
+        >>> print(text_features.shape)
+    """
+
+    def __init__(self, size, device):
+        """
+        Initialize the CLIP text encoder.
+
+        This class implements the TextModel interface using OpenAI's CLIP model for text encoding. It loads
+        a pre-trained CLIP model of the specified size and prepares it for text encoding tasks.
+
+        Args:
+            size (str): Model size identifier (e.g., 'ViT-B/32').
+            device (torch.device): Device to load the model on.
+
+        Examples:
+            >>> import torch
+            >>> from ultralytics.models.sam.modules.clip import CLIP
+            >>> clip_model = CLIP("ViT-B/32", device=torch.device("cuda:0"))
+            >>> text_features = clip_model.encode_text(["a photo of a cat", "a photo of a dog"])
+        """
+        super().__init__()
+        self.model = clip.load(size, device=device)[0]
+        self.to(device)
+        self.device = device
+        self.eval()
+
+    def tokenize(self, texts):
+        """
+        Convert input texts to CLIP tokens.
+
+        Args:
+            texts (str | List[str]): Input text or list of texts to tokenize.
+
+        Returns:
+            (torch.Tensor): Tokenized text tensor with shape (batch_size, context_length) ready for model processing.
+
+        Examples:
+            >>> model = CLIP("ViT-B/32", device="cpu")
+            >>> tokens = model.tokenize("a photo of a cat")
+            >>> print(tokens.shape)  # torch.Size([1, 77])
+        """
+        return clip.tokenize(texts).to(self.device)
+
+    @smart_inference_mode()
+    def encode_text(self, texts, dtype=torch.float32):
+        """
+        Encode tokenized texts into normalized feature vectors.
+
+        This method processes tokenized text inputs through the CLIP model to generate feature vectors, which are then
+        normalized to unit length. These normalized vectors can be used for text-image similarity comparisons.
+
+        Args:
+            texts (torch.Tensor): Tokenized text inputs, typically created using the tokenize() method.
+            dtype (torch.dtype, optional): Data type for output features. Default is torch.float32.
+
+        Returns:
+            (torch.Tensor): Normalized text feature vectors with unit length (L2 norm = 1).
+
+        Examples:
+            >>> clip_model = CLIP("ViT-B/32", device="cuda")
+            >>> tokens = clip_model.tokenize(["a photo of a cat", "a photo of a dog"])
+            >>> features = clip_model.encode_text(tokens)
+            >>> features.shape
+            torch.Size([2, 512])
+        """
+        txt_feats = self.model.encode_text(texts).to(dtype)
+        txt_feats = txt_feats / txt_feats.norm(p=2, dim=-1, keepdim=True)
+        return txt_feats
+
+
+class MobileCLIP(TextModel):
+    """
+    Implement Apple's MobileCLIP text encoder for efficient text encoding.
+
+    This class implements the TextModel interface using Apple's MobileCLIP model, providing efficient text encoding
+    capabilities for vision-language tasks.
+
+    Attributes:
+        model (mobileclip.model.MobileCLIP): The loaded MobileCLIP model.
+        tokenizer (callable): Tokenizer function for processing text inputs.
+        device (torch.device): Device where the model is loaded.
+        config_size_map (dict): Mapping from size identifiers to model configuration names.
+
+    Methods:
+        tokenize: Convert input texts to MobileCLIP tokens.
+        encode_text: Encode tokenized texts into normalized feature vectors.
+
+    Examples:
+        >>> device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        >>> text_encoder = MobileCLIP(size="s0", device=device)
+        >>> tokens = text_encoder.tokenize(["a photo of a cat", "a photo of a dog"])
+        >>> features = text_encoder.encode_text(tokens)
+    """
+
+    config_size_map = {"s0": "s0", "s1": "s1", "s2": "s2", "b": "b", "blt": "b"}
+
+    def __init__(self, size, device):
+        """
+        Initialize the MobileCLIP text encoder.
+
+        This class implements the TextModel interface using Apple's MobileCLIP model for efficient text encoding.
+
+        Args:
+            size (str): Model size identifier (e.g., 's0', 's1', 's2', 'b', 'blt').
+            device (torch.device): Device to load the model on.
+
+        Examples:
+            >>> from ultralytics.nn.modules import MobileCLIP
+            >>> import torch
+            >>> model = MobileCLIP("s0", device=torch.device("cpu"))
+            >>> tokens = model.tokenize(["a photo of a cat", "a photo of a dog"])
+            >>> features = model.encode_text(tokens)
+        """
+        try:
+            import warnings
+
+            # Suppress 'timm.models.layers is deprecated, please import via timm.layers' warning from mobileclip usage
+            with warnings.catch_warnings():
+                warnings.filterwarnings("ignore", category=FutureWarning)
+                import mobileclip
+        except ImportError:
+            # Ultralytics fork preferred since Apple MobileCLIP repo has incorrect version of torchvision
+            checks.check_requirements("git+https://github.com/ultralytics/mobileclip.git")
+            import mobileclip
+
+        super().__init__()
+        config = self.config_size_map[size]
+        file = f"mobileclip_{size}.pt"
+        if not Path(file).is_file():
+            from ultralytics import download
+
+            download(f"https://docs-assets.developer.apple.com/ml-research/datasets/mobileclip/{file}")
+        self.model = mobileclip.create_model_and_transforms(f"mobileclip_{config}", pretrained=file, device=device)[0]
+        self.tokenizer = mobileclip.get_tokenizer(f"mobileclip_{config}")
+        self.to(device)
+        self.device = device
+        self.eval()
+
+    def tokenize(self, texts):
+        """
+        Convert input texts to MobileCLIP tokens.
+
+        Args:
+            texts (list[str]): List of text strings to tokenize.
+
+        Returns:
+            (torch.Tensor): Tokenized text inputs with shape (batch_size, sequence_length).
+
+        Examples:
+            >>> model = MobileCLIP("s0", "cpu")
+            >>> tokens = model.tokenize(["a photo of a cat", "a photo of a dog"])
+        """
+        return self.tokenizer(texts).to(self.device)
+
+    @smart_inference_mode()
+    def encode_text(self, texts, dtype=torch.float32):
+        """
+        Encode tokenized texts into normalized feature vectors.
+
+        Args:
+            texts (torch.Tensor): Tokenized text inputs.
+            dtype (torch.dtype, optional): Data type for output features.
+
+        Returns:
+            (torch.Tensor): Normalized text feature vectors with L2 normalization applied.
+
+        Examples:
+            >>> model = MobileCLIP("s0", device="cpu")
+            >>> tokens = model.tokenize(["a photo of a cat", "a photo of a dog"])
+            >>> features = model.encode_text(tokens)
+            >>> features.shape
+            torch.Size([2, 512])  # Actual dimension depends on model size
+        """
+        text_features = self.model.encode_text(texts).to(dtype)
+        text_features /= text_features.norm(p=2, dim=-1, keepdim=True)
+        return text_features
+
+
+class MobileCLIPTS(TextModel):
+    """
+    Load a TorchScript traced version of MobileCLIP.
+
+    This class implements the TextModel interface using Apple's MobileCLIP model, providing efficient text encoding
+    capabilities for vision-language tasks.
+
+    Attributes:
+        encoder (mobileclip.model.MobileCLIP): The loaded MobileCLIP text encoder.
+        tokenizer (callable): Tokenizer function for processing text inputs.
+        device (torch.device): Device where the model is loaded.
+
+    Methods:
+        tokenize: Convert input texts to MobileCLIP tokens.
+        encode_text: Encode tokenized texts into normalized feature vectors.
+
+    Examples:
+        >>> device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        >>> text_encoder = MobileCLIP(device=device)
+        >>> tokens = text_encoder.tokenize(["a photo of a cat", "a photo of a dog"])
+        >>> features = text_encoder.encode_text(tokens)
+    """
+
+    def __init__(self, device):
+        """
+        Initialize the MobileCLIP text encoder.
+
+        This class implements the TextModel interface using Apple's MobileCLIP model for efficient text encoding.
+
+        Args:
+            device (torch.device): Device to load the model on.
+
+        Examples:
+            >>> from ultralytics.nn.modules import MobileCLIP
+            >>> import torch
+            >>> model = MobileCLIP(device=torch.device("cpu"))
+            >>> tokens = model.tokenize(["a photo of a cat", "a photo of a dog"])
+            >>> features = model.encode_text(tokens)
+        """
+        super().__init__()
+        from ultralytics.utils.downloads import attempt_download_asset
+
+        self.encoder = torch.jit.load(attempt_download_asset("mobileclip_blt.ts"), map_location=device)
+        self.tokenizer = clip.clip.tokenize
+        self.device = device
+
+    def tokenize(self, texts):
+        """
+        Convert input texts to MobileCLIP tokens.
+
+        Args:
+            texts (list[str]): List of text strings to tokenize.
+
+        Returns:
+            (torch.Tensor): Tokenized text inputs with shape (batch_size, sequence_length).
+
+        Examples:
+            >>> model = MobileCLIP("cpu")
+            >>> tokens = model.tokenize(["a photo of a cat", "a photo of a dog"])
+        """
+        return self.tokenizer(texts).to(self.device)
+
+    @smart_inference_mode()
+    def encode_text(self, texts, dtype=torch.float32):
+        """
+        Encode tokenized texts into normalized feature vectors.
+
+        Args:
+            texts (torch.Tensor): Tokenized text inputs.
+            dtype (torch.dtype, optional): Data type for output features.
+
+        Returns:
+            (torch.Tensor): Normalized text feature vectors with L2 normalization applied.
+
+        Examples:
+            >>> model = MobileCLIP(device="cpu")
+            >>> tokens = model.tokenize(["a photo of a cat", "a photo of a dog"])
+            >>> features = model.encode_text(tokens)
+            >>> features.shape
+            torch.Size([2, 512])  # Actual dimension depends on model size
+        """
+        return self.encoder(texts)
+
+
+def build_text_model(variant, device=None):
+    """
+    Build a text encoding model based on the specified variant.
+
+    Args:
+        variant (str): Model variant in format "base:size" (e.g., "clip:ViT-B/32" or "mobileclip:s0").
+        device (torch.device, optional): Device to load the model on.
+
+    Returns:
+        (TextModel): Instantiated text encoding model.
+
+    Examples:
+        >>> model = build_text_model("clip:ViT-B/32", device=torch.device("cuda"))
+        >>> model = build_text_model("mobileclip:s0", device=torch.device("cpu"))
+    """
+    base, size = variant.split(":")
+    if base == "clip":
+        return CLIP(size, device)
+    elif base == "mobileclip":
+        return MobileCLIPTS(device)
+    else:
+        raise ValueError(f"Unrecognized base model: '{base}'. Supported base models: 'clip', 'mobileclip'.")

ultralytics/solutions/__init__.py

@@ -0,0 +1,41 @@
+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from .ai_gym import AIGym
+from .analytics import Analytics
+from .distance_calculation import DistanceCalculation
+from .heatmap import Heatmap
+from .instance_segmentation import InstanceSegmentation
+from .object_blurrer import ObjectBlurrer
+from .object_counter import ObjectCounter
+from .object_cropper import ObjectCropper
+from .parking_management import ParkingManagement, ParkingPtsSelection
+from .queue_management import QueueManager
+from .region_counter import RegionCounter
+from .security_alarm import SecurityAlarm
+from .similarity_search import SearchApp, VisualAISearch
+from .speed_estimation import SpeedEstimator
+from .streamlit_inference import Inference
+from .trackzone import TrackZone
+from .vision_eye import VisionEye
+
+__all__ = (
+    "ObjectCounter",
+    "ObjectCropper",
+    "ObjectBlurrer",
+    "AIGym",
+    "RegionCounter",
+    "SecurityAlarm",
+    "Heatmap",
+    "InstanceSegmentation",
+    "VisionEye",
+    "SpeedEstimator",
+    "DistanceCalculation",
+    "QueueManager",
+    "ParkingManagement",
+    "ParkingPtsSelection",
+    "Analytics",
+    "Inference",
+    "TrackZone",
+    "SearchApp",
+    "VisualAISearch",
+)

ultralytics/solutions/ai_gym.py

@@ -0,0 +1,116 @@
+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+from collections import defaultdict
+
+from ultralytics.solutions.solutions import BaseSolution, SolutionAnnotator, SolutionResults
+
+
+class AIGym(BaseSolution):
+    """
+    A class to manage gym steps of people in a real-time video stream based on their poses.
+
+    This class extends BaseSolution to monitor workouts using YOLO pose estimation models. It tracks and counts
+    repetitions of exercises based on predefined angle thresholds for up and down positions.
+
+    Attributes:
+        states (Dict[float, int, str]): Stores per-track angle, count, and stage for workout monitoring.
+        up_angle (float): Angle threshold for considering the 'up' position of an exercise.
+        down_angle (float): Angle threshold for considering the 'down' position of an exercise.
+        kpts (List[int]): Indices of keypoints used for angle calculation.
+
+    Methods:
+        process: Processes a frame to detect poses, calculate angles, and count repetitions.
+
+    Examples:
+        >>> gym = AIGym(model="yolo11n-pose.pt")
+        >>> image = cv2.imread("gym_scene.jpg")
+        >>> results = gym.process(image)
+        >>> processed_image = results.plot_im
+        >>> cv2.imshow("Processed Image", processed_image)
+        >>> cv2.waitKey(0)
+    """
+
+    def __init__(self, **kwargs):
+        """
+        Initialize AIGym for workout monitoring using pose estimation and predefined angles.
+
+        Args:
+            **kwargs (Any): Keyword arguments passed to the parent class constructor.
+                model (str): Model name or path, defaults to "yolo11n-pose.pt".
+        """
+        kwargs["model"] = kwargs.get("model", "yolo11n-pose.pt")
+        super().__init__(**kwargs)
+        self.states = defaultdict(lambda: {"angle": 0, "count": 0, "stage": "-"})  # Dict for count, angle and stage
+
+        # Extract details from CFG single time for usage later
+        self.up_angle = float(self.CFG["up_angle"])  # Pose up predefined angle to consider up pose
+        self.down_angle = float(self.CFG["down_angle"])  # Pose down predefined angle to consider down pose
+        self.kpts = self.CFG["kpts"]  # User selected kpts of workouts storage for further usage
+
+    def process(self, im0):
+        """
+        Monitor workouts using Ultralytics YOLO Pose Model.
+
+        This function processes an input image to track and analyze human poses for workout monitoring. It uses
+        the YOLO Pose model to detect keypoints, estimate angles, and count repetitions based on predefined
+        angle thresholds.
+
+        Args:
+            im0 (np.ndarray): Input image for processing.
+
+        Returns:
+            (SolutionResults): Contains processed image `plot_im`,
+                'workout_count' (list of completed reps),
+                'workout_stage' (list of current stages),
+                'workout_angle' (list of angles), and
+                'total_tracks' (total number of tracked individuals).
+
+        Examples:
+            >>> gym = AIGym()
+            >>> image = cv2.imread("workout.jpg")
+            >>> results = gym.process(image)
+            >>> processed_image = results.plot_im
+        """
+        annotator = SolutionAnnotator(im0, line_width=self.line_width)  # Initialize annotator
+
+        self.extract_tracks(im0)  # Extract tracks (bounding boxes, classes, and masks)
+        tracks = self.tracks[0]
+
+        if tracks.boxes.id is not None:
+            track_ids = tracks.boxes.id.cpu().tolist()
+            kpt_data = tracks.keypoints.data.cpu()  # Avoid repeated .cpu() calls
+
+            for i, k in enumerate(kpt_data):
+                track_id = int(track_ids[i])  # get track id
+                state = self.states[track_id]  # get state details
+                # Get keypoints and estimate the angle
+                state["angle"] = annotator.estimate_pose_angle(*[k[int(idx)] for idx in self.kpts])
+                annotator.draw_specific_kpts(k, self.kpts, radius=self.line_width * 3)
+
+                # Determine stage and count logic based on angle thresholds
+                if state["angle"] < self.down_angle:
+                    if state["stage"] == "up":
+                        state["count"] += 1
+                    state["stage"] = "down"
+                elif state["angle"] > self.up_angle:
+                    state["stage"] = "up"
+
+                # Display angle, count, and stage text
+                if self.show_labels:
+                    annotator.plot_angle_and_count_and_stage(
+                        angle_text=state["angle"],  # angle text for display
+                        count_text=state["count"],  # count text for workouts
+                        stage_text=state["stage"],  # stage position text
+                        center_kpt=k[int(self.kpts[1])],  # center keypoint for display
+                    )
+        plot_im = annotator.result()
+        self.display_output(plot_im)  # Display output image, if environment support display
+
+        # Return SolutionResults
+        return SolutionResults(
+            plot_im=plot_im,
+            workout_count=[v["count"] for v in self.states.values()],
+            workout_stage=[v["stage"] for v in self.states.values()],
+            workout_angle=[v["angle"] for v in self.states.values()],
+            total_tracks=len(self.track_ids),
+        )