Zum Inhalt springen

Referenz fĂŒr ultralytics/models/sam/modules/encoders.py

Hinweis

Diese Datei ist verfĂŒgbar unter https://github.com/ultralytics/ ultralytics/blob/main/ ultralytics/models/ sam/modules/encoders .py. Wenn du ein Problem entdeckst, hilf bitte mit, es zu beheben, indem du einen Pull Request đŸ› ïž einreichst. Vielen Dank 🙏!



ultralytics.models.sam.modules.encoders.ImageEncoderViT

Basen: Module

Ein Bildkodierer, der die Architektur des Vision Transformer (ViT) nutzt, um ein Bild in einen kompakten latenten Raum zu kodieren. Der Kodierer nimmt ein Bild, unterteilt es in Teilbilder und verarbeitet diese Teilbilder durch eine Reihe von Transformatorblöcken. Die kodierten Patches werden dann durch einen Hals verarbeitet, um die endgĂŒltige kodierte Darstellung zu erzeugen.

Diese Klasse und ihre unterstĂŒtzenden Funktionen wurden leicht an das ViTDet-Backbone angepasst, das unter https://github.com/facebookresearch/detectron2/blob/main/detectron2/modeling/backbone/vit.py.

Attribute:

Name Typ Beschreibung
img_size int

Dimension der Eingabebilder, die als quadratisch angenommen wird.

patch_embed PatchEmbed

Modul zum Einbetten von Patches.

pos_embed Parameter

Absolute positionale Einbettung fĂŒr Patches.

blocks ModuleList

Liste der Transformatorblöcke fĂŒr die Verarbeitung von Patch-Embeddings.

neck Sequential

Neck-Modul, um die Ausgabe weiter zu verarbeiten.

Quellcode in ultralytics/models/sam/modules/encoders.py
class ImageEncoderViT(nn.Module):
    """
    An image encoder using Vision Transformer (ViT) architecture for encoding an image into a compact latent space. The
    encoder takes an image, splits it into patches, and processes these patches through a series of transformer blocks.
    The encoded patches are then processed through a neck to generate the final encoded representation.

    This class and its supporting functions below lightly adapted from the ViTDet backbone available at
    https://github.com/facebookresearch/detectron2/blob/main/detectron2/modeling/backbone/vit.py.

    Attributes:
        img_size (int): Dimension of input images, assumed to be square.
        patch_embed (PatchEmbed): Module for patch embedding.
        pos_embed (nn.Parameter, optional): Absolute positional embedding for patches.
        blocks (nn.ModuleList): List of transformer blocks for processing patch embeddings.
        neck (nn.Sequential): Neck module to further process the output.
    """

    def __init__(
        self,
        img_size: int = 1024,
        patch_size: int = 16,
        in_chans: int = 3,
        embed_dim: int = 768,
        depth: int = 12,
        num_heads: int = 12,
        mlp_ratio: float = 4.0,
        out_chans: int = 256,
        qkv_bias: bool = True,
        norm_layer: Type[nn.Module] = nn.LayerNorm,
        act_layer: Type[nn.Module] = nn.GELU,
        use_abs_pos: bool = True,
        use_rel_pos: bool = False,
        rel_pos_zero_init: bool = True,
        window_size: int = 0,
        global_attn_indexes: Tuple[int, ...] = (),
    ) -> None:
        """
        Args:
            img_size (int): Input image size.
            patch_size (int): Patch size.
            in_chans (int): Number of input image channels.
            embed_dim (int): Patch embedding dimension.
            depth (int): Depth of ViT.
            num_heads (int): Number of attention heads in each ViT block.
            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
            qkv_bias (bool): If True, add a learnable bias to query, key, value.
            norm_layer (nn.Module): Normalization layer.
            act_layer (nn.Module): Activation layer.
            use_abs_pos (bool): If True, use absolute positional embeddings.
            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
            window_size (int): Window size for window attention blocks.
            global_attn_indexes (list): Indexes for blocks using global attention.
        """
        super().__init__()
        self.img_size = img_size

        self.patch_embed = PatchEmbed(
            kernel_size=(patch_size, patch_size),
            stride=(patch_size, patch_size),
            in_chans=in_chans,
            embed_dim=embed_dim,
        )

        self.pos_embed: Optional[nn.Parameter] = None
        if use_abs_pos:
            # Initialize absolute positional embedding with pretrain image size.
            self.pos_embed = nn.Parameter(torch.zeros(1, img_size // patch_size, img_size // patch_size, embed_dim))

        self.blocks = nn.ModuleList()
        for i in range(depth):
            block = Block(
                dim=embed_dim,
                num_heads=num_heads,
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                norm_layer=norm_layer,
                act_layer=act_layer,
                use_rel_pos=use_rel_pos,
                rel_pos_zero_init=rel_pos_zero_init,
                window_size=window_size if i not in global_attn_indexes else 0,
                input_size=(img_size // patch_size, img_size // patch_size),
            )
            self.blocks.append(block)

        self.neck = nn.Sequential(
            nn.Conv2d(
                embed_dim,
                out_chans,
                kernel_size=1,
                bias=False,
            ),
            LayerNorm2d(out_chans),
            nn.Conv2d(
                out_chans,
                out_chans,
                kernel_size=3,
                padding=1,
                bias=False,
            ),
            LayerNorm2d(out_chans),
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        """Processes input through patch embedding, applies positional embedding if present, and passes through blocks
        and neck.
        """
        x = self.patch_embed(x)
        if self.pos_embed is not None:
            x = x + self.pos_embed
        for blk in self.blocks:
            x = blk(x)
        return self.neck(x.permute(0, 3, 1, 2))

__init__(img_size=1024, patch_size=16, in_chans=3, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4.0, out_chans=256, qkv_bias=True, norm_layer=nn.LayerNorm, act_layer=nn.GELU, use_abs_pos=True, use_rel_pos=False, rel_pos_zero_init=True, window_size=0, global_attn_indexes=())

Parameter:

Name Typ Beschreibung Standard
img_size int

GrĂ¶ĂŸe des Eingangsbildes.

1024
patch_size int

Patch-GrĂ¶ĂŸe.

16
in_chans int

Anzahl der EingangsbildkanÀle.

3
embed_dim int

Patch-Einbettungsdimension.

768
depth int

Tiefe des ViT.

12
num_heads int

Anzahl der Aufmerksamkeitsköpfe in jedem ViT-Block.

12
mlp_ratio float

VerhÀltnis von mlp hidden dim zu embedding dim.

4.0
qkv_bias bool

Wenn True, fĂŒge einen lernbaren Bias zu query, key, value hinzu.

True
norm_layer Module

Normalisierungsschicht.

LayerNorm
act_layer Module

Aktivierungsschicht.

GELU
use_abs_pos bool

Wenn True, werden absolute Positionseinbettungen verwendet.

True
use_rel_pos bool

Wenn "True", fĂŒge der Aufmerksamkeitskarte relative Positionseinbettungen hinzu.

False
rel_pos_zero_init bool

Wenn True, werden die relativen Positionsparameter mit Null initialisiert.

True
window_size int

FenstergrĂ¶ĂŸe fĂŒr Aufmerksamkeitsblöcke.

0
global_attn_indexes list

Indizes fĂŒr Blöcke mit globaler Aufmerksamkeit.

()
Quellcode in ultralytics/models/sam/modules/encoders.py
def __init__(
    self,
    img_size: int = 1024,
    patch_size: int = 16,
    in_chans: int = 3,
    embed_dim: int = 768,
    depth: int = 12,
    num_heads: int = 12,
    mlp_ratio: float = 4.0,
    out_chans: int = 256,
    qkv_bias: bool = True,
    norm_layer: Type[nn.Module] = nn.LayerNorm,
    act_layer: Type[nn.Module] = nn.GELU,
    use_abs_pos: bool = True,
    use_rel_pos: bool = False,
    rel_pos_zero_init: bool = True,
    window_size: int = 0,
    global_attn_indexes: Tuple[int, ...] = (),
) -> None:
    """
    Args:
        img_size (int): Input image size.
        patch_size (int): Patch size.
        in_chans (int): Number of input image channels.
        embed_dim (int): Patch embedding dimension.
        depth (int): Depth of ViT.
        num_heads (int): Number of attention heads in each ViT block.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool): If True, add a learnable bias to query, key, value.
        norm_layer (nn.Module): Normalization layer.
        act_layer (nn.Module): Activation layer.
        use_abs_pos (bool): If True, use absolute positional embeddings.
        use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
        rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
        window_size (int): Window size for window attention blocks.
        global_attn_indexes (list): Indexes for blocks using global attention.
    """
    super().__init__()
    self.img_size = img_size

    self.patch_embed = PatchEmbed(
        kernel_size=(patch_size, patch_size),
        stride=(patch_size, patch_size),
        in_chans=in_chans,
        embed_dim=embed_dim,
    )

    self.pos_embed: Optional[nn.Parameter] = None
    if use_abs_pos:
        # Initialize absolute positional embedding with pretrain image size.
        self.pos_embed = nn.Parameter(torch.zeros(1, img_size // patch_size, img_size // patch_size, embed_dim))

    self.blocks = nn.ModuleList()
    for i in range(depth):
        block = Block(
            dim=embed_dim,
            num_heads=num_heads,
            mlp_ratio=mlp_ratio,
            qkv_bias=qkv_bias,
            norm_layer=norm_layer,
            act_layer=act_layer,
            use_rel_pos=use_rel_pos,
            rel_pos_zero_init=rel_pos_zero_init,
            window_size=window_size if i not in global_attn_indexes else 0,
            input_size=(img_size // patch_size, img_size // patch_size),
        )
        self.blocks.append(block)

    self.neck = nn.Sequential(
        nn.Conv2d(
            embed_dim,
            out_chans,
            kernel_size=1,
            bias=False,
        ),
        LayerNorm2d(out_chans),
        nn.Conv2d(
            out_chans,
            out_chans,
            kernel_size=3,
            padding=1,
            bias=False,
        ),
        LayerNorm2d(out_chans),
    )

forward(x)

Verarbeitet die Eingabe durch Patch Embedding, wendet, falls vorhanden, Positional Embedding an und durchlÀuft Blöcke und Hals.

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward(self, x: torch.Tensor) -> torch.Tensor:
    """Processes input through patch embedding, applies positional embedding if present, and passes through blocks
    and neck.
    """
    x = self.patch_embed(x)
    if self.pos_embed is not None:
        x = x + self.pos_embed
    for blk in self.blocks:
        x = blk(x)
    return self.neck(x.permute(0, 3, 1, 2))



ultralytics.models.sam.modules.encoders.PromptEncoder

Basen: Module

Kodiert verschiedene Arten von Aufforderungen, darunter Punkte, KĂ€stchen und Masken, fĂŒr die Eingabe in den Maskendekoder von SAM. Der Kodierer erzeugt sowohl spĂ€rliche als auch dichte Einbettungen fĂŒr die Eingabeaufforderungen.

Attribute:

Name Typ Beschreibung
embed_dim int

Dimension der Einbettungen.

input_image_size Tuple[int, int]

GrĂ¶ĂŸe des Eingangsbildes als (H, W).

image_embedding_size Tuple[int, int]

RĂ€umliche GrĂ¶ĂŸe der Bildeinbettung als (H, W).

pe_layer PositionEmbeddingRandom

Modul fĂŒr die Einbettung von zufĂ€lligen Positionen.

num_point_embeddings int

Anzahl der Punkteinbettungen fĂŒr verschiedene Arten von Punkten.

point_embeddings ModuleList

Liste der Punkteinbettungen.

not_a_point_embed Embedding

Einbettung fĂŒr Punkte, die nicht Teil eines Labels sind.

mask_input_size Tuple[int, int]

GrĂ¶ĂŸe der Eingabemaske.

mask_downscaling Sequential

Neuronales Netzwerk fĂŒr das Downscaling der Maske.

no_mask_embed Embedding

Einbettung fĂŒr FĂ€lle, in denen keine Maske vorgesehen ist.

Quellcode in ultralytics/models/sam/modules/encoders.py
class PromptEncoder(nn.Module):
    """
    Encodes different types of prompts, including points, boxes, and masks, for input to SAM's mask decoder. The encoder
    produces both sparse and dense embeddings for the input prompts.

    Attributes:
        embed_dim (int): Dimension of the embeddings.
        input_image_size (Tuple[int, int]): Size of the input image as (H, W).
        image_embedding_size (Tuple[int, int]): Spatial size of the image embedding as (H, W).
        pe_layer (PositionEmbeddingRandom): Module for random position embedding.
        num_point_embeddings (int): Number of point embeddings for different types of points.
        point_embeddings (nn.ModuleList): List of point embeddings.
        not_a_point_embed (nn.Embedding): Embedding for points that are not a part of any label.
        mask_input_size (Tuple[int, int]): Size of the input mask.
        mask_downscaling (nn.Sequential): Neural network for downscaling the mask.
        no_mask_embed (nn.Embedding): Embedding for cases where no mask is provided.
    """

    def __init__(
        self,
        embed_dim: int,
        image_embedding_size: Tuple[int, int],
        input_image_size: Tuple[int, int],
        mask_in_chans: int,
        activation: Type[nn.Module] = nn.GELU,
    ) -> None:
        """
        Encodes prompts for input to SAM's mask decoder.

        Args:
          embed_dim (int): The prompts' embedding dimension
          image_embedding_size (tuple(int, int)): The spatial size of the
            image embedding, as (H, W).
          input_image_size (int): The padded size of the image as input
            to the image encoder, as (H, W).
          mask_in_chans (int): The number of hidden channels used for
            encoding input masks.
          activation (nn.Module): The activation to use when encoding
            input masks.
        """
        super().__init__()
        self.embed_dim = embed_dim
        self.input_image_size = input_image_size
        self.image_embedding_size = image_embedding_size
        self.pe_layer = PositionEmbeddingRandom(embed_dim // 2)

        self.num_point_embeddings: int = 4  # pos/neg point + 2 box corners
        point_embeddings = [nn.Embedding(1, embed_dim) for _ in range(self.num_point_embeddings)]
        self.point_embeddings = nn.ModuleList(point_embeddings)
        self.not_a_point_embed = nn.Embedding(1, embed_dim)

        self.mask_input_size = (4 * image_embedding_size[0], 4 * image_embedding_size[1])
        self.mask_downscaling = nn.Sequential(
            nn.Conv2d(1, mask_in_chans // 4, kernel_size=2, stride=2),
            LayerNorm2d(mask_in_chans // 4),
            activation(),
            nn.Conv2d(mask_in_chans // 4, mask_in_chans, kernel_size=2, stride=2),
            LayerNorm2d(mask_in_chans),
            activation(),
            nn.Conv2d(mask_in_chans, embed_dim, kernel_size=1),
        )
        self.no_mask_embed = nn.Embedding(1, embed_dim)

    def get_dense_pe(self) -> torch.Tensor:
        """
        Returns the positional encoding used to encode point prompts, applied to a dense set of points the shape of the
        image encoding.

        Returns:
          torch.Tensor: Positional encoding with shape 1x(embed_dim)x(embedding_h)x(embedding_w)
        """
        return self.pe_layer(self.image_embedding_size).unsqueeze(0)

    def _embed_points(self, points: torch.Tensor, labels: torch.Tensor, pad: bool) -> torch.Tensor:
        """Embeds point prompts."""
        points = points + 0.5  # Shift to center of pixel
        if pad:
            padding_point = torch.zeros((points.shape[0], 1, 2), device=points.device)
            padding_label = -torch.ones((labels.shape[0], 1), device=labels.device)
            points = torch.cat([points, padding_point], dim=1)
            labels = torch.cat([labels, padding_label], dim=1)
        point_embedding = self.pe_layer.forward_with_coords(points, self.input_image_size)
        point_embedding[labels == -1] = 0.0
        point_embedding[labels == -1] += self.not_a_point_embed.weight
        point_embedding[labels == 0] += self.point_embeddings[0].weight
        point_embedding[labels == 1] += self.point_embeddings[1].weight
        return point_embedding

    def _embed_boxes(self, boxes: torch.Tensor) -> torch.Tensor:
        """Embeds box prompts."""
        boxes = boxes + 0.5  # Shift to center of pixel
        coords = boxes.reshape(-1, 2, 2)
        corner_embedding = self.pe_layer.forward_with_coords(coords, self.input_image_size)
        corner_embedding[:, 0, :] += self.point_embeddings[2].weight
        corner_embedding[:, 1, :] += self.point_embeddings[3].weight
        return corner_embedding

    def _embed_masks(self, masks: torch.Tensor) -> torch.Tensor:
        """Embeds mask inputs."""
        return self.mask_downscaling(masks)

    def _get_batch_size(
        self,
        points: Optional[Tuple[torch.Tensor, torch.Tensor]],
        boxes: Optional[torch.Tensor],
        masks: Optional[torch.Tensor],
    ) -> int:
        """Gets the batch size of the output given the batch size of the input prompts."""
        if points is not None:
            return points[0].shape[0]
        elif boxes is not None:
            return boxes.shape[0]
        elif masks is not None:
            return masks.shape[0]
        else:
            return 1

    def _get_device(self) -> torch.device:
        """Returns the device of the first point embedding's weight tensor."""
        return self.point_embeddings[0].weight.device

    def forward(
        self,
        points: Optional[Tuple[torch.Tensor, torch.Tensor]],
        boxes: Optional[torch.Tensor],
        masks: Optional[torch.Tensor],
    ) -> Tuple[torch.Tensor, torch.Tensor]:
        """
        Embeds different types of prompts, returning both sparse and dense embeddings.

        Args:
          points (tuple(torch.Tensor, torch.Tensor), None): point coordinates and labels to embed.
          boxes (torch.Tensor, None): boxes to embed
          masks (torch.Tensor, None): masks to embed

        Returns:
          torch.Tensor: sparse embeddings for the points and boxes, with shape BxNx(embed_dim), where N is determined
            by the number of input points and boxes.
          torch.Tensor: dense embeddings for the masks, in the shape Bx(embed_dim)x(embed_H)x(embed_W)
        """
        bs = self._get_batch_size(points, boxes, masks)
        sparse_embeddings = torch.empty((bs, 0, self.embed_dim), device=self._get_device())
        if points is not None:
            coords, labels = points
            point_embeddings = self._embed_points(coords, labels, pad=(boxes is None))
            sparse_embeddings = torch.cat([sparse_embeddings, point_embeddings], dim=1)
        if boxes is not None:
            box_embeddings = self._embed_boxes(boxes)
            sparse_embeddings = torch.cat([sparse_embeddings, box_embeddings], dim=1)

        if masks is not None:
            dense_embeddings = self._embed_masks(masks)
        else:
            dense_embeddings = self.no_mask_embed.weight.reshape(1, -1, 1, 1).expand(
                bs, -1, self.image_embedding_size[0], self.image_embedding_size[1]
            )

        return sparse_embeddings, dense_embeddings

__init__(embed_dim, image_embedding_size, input_image_size, mask_in_chans, activation=nn.GELU)

VerschlĂŒsselt Aufforderungen zur Eingabe in den Maskendecoder von SAM.

Parameter:

Name Typ Beschreibung Standard
embed_dim int

Die Dimension der Einbettung der Prompts

erforderlich
image_embedding_size tuple(int, int

Die rĂ€umliche GrĂ¶ĂŸe der der Bildeinbettung, als (H, W).

erforderlich
input_image_size int

Die aufgefĂŒllte GrĂ¶ĂŸe des Bildes als Eingabe an den Bildkodierer, als (H, W).

erforderlich
mask_in_chans int

Die Anzahl der versteckten KanĂ€le, die fĂŒr die Codierung der Eingabemasken verwendet werden.

erforderlich
activation Module

Die Aktivierung, die bei der Kodierung von Eingabemasken.

GELU
Quellcode in ultralytics/models/sam/modules/encoders.py
def __init__(
    self,
    embed_dim: int,
    image_embedding_size: Tuple[int, int],
    input_image_size: Tuple[int, int],
    mask_in_chans: int,
    activation: Type[nn.Module] = nn.GELU,
) -> None:
    """
    Encodes prompts for input to SAM's mask decoder.

    Args:
      embed_dim (int): The prompts' embedding dimension
      image_embedding_size (tuple(int, int)): The spatial size of the
        image embedding, as (H, W).
      input_image_size (int): The padded size of the image as input
        to the image encoder, as (H, W).
      mask_in_chans (int): The number of hidden channels used for
        encoding input masks.
      activation (nn.Module): The activation to use when encoding
        input masks.
    """
    super().__init__()
    self.embed_dim = embed_dim
    self.input_image_size = input_image_size
    self.image_embedding_size = image_embedding_size
    self.pe_layer = PositionEmbeddingRandom(embed_dim // 2)

    self.num_point_embeddings: int = 4  # pos/neg point + 2 box corners
    point_embeddings = [nn.Embedding(1, embed_dim) for _ in range(self.num_point_embeddings)]
    self.point_embeddings = nn.ModuleList(point_embeddings)
    self.not_a_point_embed = nn.Embedding(1, embed_dim)

    self.mask_input_size = (4 * image_embedding_size[0], 4 * image_embedding_size[1])
    self.mask_downscaling = nn.Sequential(
        nn.Conv2d(1, mask_in_chans // 4, kernel_size=2, stride=2),
        LayerNorm2d(mask_in_chans // 4),
        activation(),
        nn.Conv2d(mask_in_chans // 4, mask_in_chans, kernel_size=2, stride=2),
        LayerNorm2d(mask_in_chans),
        activation(),
        nn.Conv2d(mask_in_chans, embed_dim, kernel_size=1),
    )
    self.no_mask_embed = nn.Embedding(1, embed_dim)

forward(points, boxes, masks)

Bettet verschiedene Arten von Aufforderungen ein und liefert sowohl spÀrliche als auch dichte Einbettungen.

Parameter:

Name Typ Beschreibung Standard
points (tuple(Tensor, Tensor), None)

Punktkoordinaten und Beschriftungen zum Einbetten.

erforderlich
boxes (Tensor, None)

Boxen zum Einbetten

erforderlich
masks (Tensor, None)

Masken zum Einbetten

erforderlich

Retouren:

Typ Beschreibung
Tensor

torch.Tensor: Sparse Embeddings fĂŒr die Punkte und Boxen, mit der Form BxNx(embed_dim), wobei N durch die Anzahl der durch die Anzahl der eingegebenen Punkte und KĂ€stchen bestimmt wird.

Tensor

torch.Tensor: dichte Einbettungen fĂŒr die Masken, in der Form Bx(embed_dim)x(embed_H)x(embed_W)

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward(
    self,
    points: Optional[Tuple[torch.Tensor, torch.Tensor]],
    boxes: Optional[torch.Tensor],
    masks: Optional[torch.Tensor],
) -> Tuple[torch.Tensor, torch.Tensor]:
    """
    Embeds different types of prompts, returning both sparse and dense embeddings.

    Args:
      points (tuple(torch.Tensor, torch.Tensor), None): point coordinates and labels to embed.
      boxes (torch.Tensor, None): boxes to embed
      masks (torch.Tensor, None): masks to embed

    Returns:
      torch.Tensor: sparse embeddings for the points and boxes, with shape BxNx(embed_dim), where N is determined
        by the number of input points and boxes.
      torch.Tensor: dense embeddings for the masks, in the shape Bx(embed_dim)x(embed_H)x(embed_W)
    """
    bs = self._get_batch_size(points, boxes, masks)
    sparse_embeddings = torch.empty((bs, 0, self.embed_dim), device=self._get_device())
    if points is not None:
        coords, labels = points
        point_embeddings = self._embed_points(coords, labels, pad=(boxes is None))
        sparse_embeddings = torch.cat([sparse_embeddings, point_embeddings], dim=1)
    if boxes is not None:
        box_embeddings = self._embed_boxes(boxes)
        sparse_embeddings = torch.cat([sparse_embeddings, box_embeddings], dim=1)

    if masks is not None:
        dense_embeddings = self._embed_masks(masks)
    else:
        dense_embeddings = self.no_mask_embed.weight.reshape(1, -1, 1, 1).expand(
            bs, -1, self.image_embedding_size[0], self.image_embedding_size[1]
        )

    return sparse_embeddings, dense_embeddings

get_dense_pe()

Gibt die Positionskodierung zurĂŒck, die fĂŒr die Kodierung von Punktaufforderungen verwendet wird, angewendet auf eine dichte Menge von Punkten, die die Form der Bildkodierung.

Retouren:

Typ Beschreibung
Tensor

torch.Tensor: Positionskodierung mit der Form 1x(embed_dim)x(embedding_h)x(embedding_w)

Quellcode in ultralytics/models/sam/modules/encoders.py
def get_dense_pe(self) -> torch.Tensor:
    """
    Returns the positional encoding used to encode point prompts, applied to a dense set of points the shape of the
    image encoding.

    Returns:
      torch.Tensor: Positional encoding with shape 1x(embed_dim)x(embedding_h)x(embedding_w)
    """
    return self.pe_layer(self.image_embedding_size).unsqueeze(0)



ultralytics.models.sam.modules.encoders.PositionEmbeddingRandom

Basen: Module

Positionelle Kodierung mit zufÀlligen rÀumlichen Frequenzen.

Quellcode in ultralytics/models/sam/modules/encoders.py
class PositionEmbeddingRandom(nn.Module):
    """Positional encoding using random spatial frequencies."""

    def __init__(self, num_pos_feats: int = 64, scale: Optional[float] = None) -> None:
        """Initializes a position embedding using random spatial frequencies."""
        super().__init__()
        if scale is None or scale <= 0.0:
            scale = 1.0
        self.register_buffer("positional_encoding_gaussian_matrix", scale * torch.randn((2, num_pos_feats)))

        # Set non-deterministic for forward() error 'cumsum_cuda_kernel does not have a deterministic implementation'
        torch.use_deterministic_algorithms(False)
        torch.backends.cudnn.deterministic = False

    def _pe_encoding(self, coords: torch.Tensor) -> torch.Tensor:
        """Positionally encode points that are normalized to [0,1]."""
        # Assuming coords are in [0, 1]^2 square and have d_1 x ... x d_n x 2 shape
        coords = 2 * coords - 1
        coords = coords @ self.positional_encoding_gaussian_matrix
        coords = 2 * np.pi * coords
        # Outputs d_1 x ... x d_n x C shape
        return torch.cat([torch.sin(coords), torch.cos(coords)], dim=-1)

    def forward(self, size: Tuple[int, int]) -> torch.Tensor:
        """Generate positional encoding for a grid of the specified size."""
        h, w = size
        device: Any = self.positional_encoding_gaussian_matrix.device
        grid = torch.ones((h, w), device=device, dtype=torch.float32)
        y_embed = grid.cumsum(dim=0) - 0.5
        x_embed = grid.cumsum(dim=1) - 0.5
        y_embed = y_embed / h
        x_embed = x_embed / w

        pe = self._pe_encoding(torch.stack([x_embed, y_embed], dim=-1))
        return pe.permute(2, 0, 1)  # C x H x W

    def forward_with_coords(self, coords_input: torch.Tensor, image_size: Tuple[int, int]) -> torch.Tensor:
        """Positionally encode points that are not normalized to [0,1]."""
        coords = coords_input.clone()
        coords[:, :, 0] = coords[:, :, 0] / image_size[1]
        coords[:, :, 1] = coords[:, :, 1] / image_size[0]
        return self._pe_encoding(coords.to(torch.float))  # B x N x C

__init__(num_pos_feats=64, scale=None)

Initialisiert eine Positionseinbettung mit zufÀlligen rÀumlichen Frequenzen.

Quellcode in ultralytics/models/sam/modules/encoders.py
def __init__(self, num_pos_feats: int = 64, scale: Optional[float] = None) -> None:
    """Initializes a position embedding using random spatial frequencies."""
    super().__init__()
    if scale is None or scale <= 0.0:
        scale = 1.0
    self.register_buffer("positional_encoding_gaussian_matrix", scale * torch.randn((2, num_pos_feats)))

    # Set non-deterministic for forward() error 'cumsum_cuda_kernel does not have a deterministic implementation'
    torch.use_deterministic_algorithms(False)
    torch.backends.cudnn.deterministic = False

forward(size)

Erstelle eine Positionskodierung fĂŒr ein Raster mit der angegebenen GrĂ¶ĂŸe.

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward(self, size: Tuple[int, int]) -> torch.Tensor:
    """Generate positional encoding for a grid of the specified size."""
    h, w = size
    device: Any = self.positional_encoding_gaussian_matrix.device
    grid = torch.ones((h, w), device=device, dtype=torch.float32)
    y_embed = grid.cumsum(dim=0) - 0.5
    x_embed = grid.cumsum(dim=1) - 0.5
    y_embed = y_embed / h
    x_embed = x_embed / w

    pe = self._pe_encoding(torch.stack([x_embed, y_embed], dim=-1))
    return pe.permute(2, 0, 1)  # C x H x W

forward_with_coords(coords_input, image_size)

Kodiere Punkte, die nicht auf [0,1] normalisiert sind, in ihrer Position.

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward_with_coords(self, coords_input: torch.Tensor, image_size: Tuple[int, int]) -> torch.Tensor:
    """Positionally encode points that are not normalized to [0,1]."""
    coords = coords_input.clone()
    coords[:, :, 0] = coords[:, :, 0] / image_size[1]
    coords[:, :, 1] = coords[:, :, 1] / image_size[0]
    return self._pe_encoding(coords.to(torch.float))  # B x N x C



ultralytics.models.sam.modules.encoders.Block

Basen: Module

Transformatorblöcke mit UnterstĂŒtzung von Fensteraufmerksamkeit und Restausbreitungsblöcken.

Quellcode in ultralytics/models/sam/modules/encoders.py
class Block(nn.Module):
    """Transformer blocks with support of window attention and residual propagation blocks."""

    def __init__(
        self,
        dim: int,
        num_heads: int,
        mlp_ratio: float = 4.0,
        qkv_bias: bool = True,
        norm_layer: Type[nn.Module] = nn.LayerNorm,
        act_layer: Type[nn.Module] = nn.GELU,
        use_rel_pos: bool = False,
        rel_pos_zero_init: bool = True,
        window_size: int = 0,
        input_size: Optional[Tuple[int, int]] = None,
    ) -> None:
        """
        Args:
            dim (int): Number of input channels.
            num_heads (int): Number of attention heads in each ViT block.
            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
            qkv_bias (bool): If True, add a learnable bias to query, key, value.
            norm_layer (nn.Module): Normalization layer.
            act_layer (nn.Module): Activation layer.
            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
            window_size (int): Window size for window attention blocks. If it equals 0, then
                use global attention.
            input_size (tuple(int, int), None): Input resolution for calculating the relative
                positional parameter size.
        """
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim,
            num_heads=num_heads,
            qkv_bias=qkv_bias,
            use_rel_pos=use_rel_pos,
            rel_pos_zero_init=rel_pos_zero_init,
            input_size=input_size if window_size == 0 else (window_size, window_size),
        )

        self.norm2 = norm_layer(dim)
        self.mlp = MLPBlock(embedding_dim=dim, mlp_dim=int(dim * mlp_ratio), act=act_layer)

        self.window_size = window_size

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        """Executes a forward pass through the transformer block with window attention and non-overlapping windows."""
        shortcut = x
        x = self.norm1(x)
        # Window partition
        if self.window_size > 0:
            H, W = x.shape[1], x.shape[2]
            x, pad_hw = window_partition(x, self.window_size)

        x = self.attn(x)
        # Reverse window partition
        if self.window_size > 0:
            x = window_unpartition(x, self.window_size, pad_hw, (H, W))

        x = shortcut + x
        return x + self.mlp(self.norm2(x))

__init__(dim, num_heads, mlp_ratio=4.0, qkv_bias=True, norm_layer=nn.LayerNorm, act_layer=nn.GELU, use_rel_pos=False, rel_pos_zero_init=True, window_size=0, input_size=None)

Parameter:

Name Typ Beschreibung Standard
dim int

Anzahl der EingangskanÀle.

erforderlich
num_heads int

Anzahl der Aufmerksamkeitsköpfe in jedem ViT-Block.

erforderlich
mlp_ratio float

VerhÀltnis von mlp hidden dim zu embedding dim.

4.0
qkv_bias bool

Wenn True, fĂŒge einen lernbaren Bias zu query, key, value hinzu.

True
norm_layer Module

Normalisierungsschicht.

LayerNorm
act_layer Module

Aktivierungsschicht.

GELU
use_rel_pos bool

Wenn "True", fĂŒge der Aufmerksamkeitskarte relative Positionseinbettungen hinzu.

False
rel_pos_zero_init bool

Wenn True, werden die relativen Positionsparameter mit Null initialisiert.

True
window_size int

FenstergrĂ¶ĂŸe fĂŒr die Aufmerksamkeitsblöcke des Fensters. Wenn sie gleich 0 ist, dann globale Aufmerksamkeit verwenden.

0
input_size (tuple(int, int), None)

Eingabeauflösung fĂŒr die Berechnung der relativen GrĂ¶ĂŸe der Positionsparameter.

None
Quellcode in ultralytics/models/sam/modules/encoders.py
def __init__(
    self,
    dim: int,
    num_heads: int,
    mlp_ratio: float = 4.0,
    qkv_bias: bool = True,
    norm_layer: Type[nn.Module] = nn.LayerNorm,
    act_layer: Type[nn.Module] = nn.GELU,
    use_rel_pos: bool = False,
    rel_pos_zero_init: bool = True,
    window_size: int = 0,
    input_size: Optional[Tuple[int, int]] = None,
) -> None:
    """
    Args:
        dim (int): Number of input channels.
        num_heads (int): Number of attention heads in each ViT block.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool): If True, add a learnable bias to query, key, value.
        norm_layer (nn.Module): Normalization layer.
        act_layer (nn.Module): Activation layer.
        use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
        rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
        window_size (int): Window size for window attention blocks. If it equals 0, then
            use global attention.
        input_size (tuple(int, int), None): Input resolution for calculating the relative
            positional parameter size.
    """
    super().__init__()
    self.norm1 = norm_layer(dim)
    self.attn = Attention(
        dim,
        num_heads=num_heads,
        qkv_bias=qkv_bias,
        use_rel_pos=use_rel_pos,
        rel_pos_zero_init=rel_pos_zero_init,
        input_size=input_size if window_size == 0 else (window_size, window_size),
    )

    self.norm2 = norm_layer(dim)
    self.mlp = MLPBlock(embedding_dim=dim, mlp_dim=int(dim * mlp_ratio), act=act_layer)

    self.window_size = window_size

forward(x)

FĂŒhrt einen VorwĂ€rtsdurchlauf durch den Trafoblock mit Fensteraufmerksamkeit und nicht ĂŒberlappenden Fenstern aus.

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward(self, x: torch.Tensor) -> torch.Tensor:
    """Executes a forward pass through the transformer block with window attention and non-overlapping windows."""
    shortcut = x
    x = self.norm1(x)
    # Window partition
    if self.window_size > 0:
        H, W = x.shape[1], x.shape[2]
        x, pad_hw = window_partition(x, self.window_size)

    x = self.attn(x)
    # Reverse window partition
    if self.window_size > 0:
        x = window_unpartition(x, self.window_size, pad_hw, (H, W))

    x = shortcut + x
    return x + self.mlp(self.norm2(x))



ultralytics.models.sam.modules.encoders.Attention

Basen: Module

Mehrköpfiger Aufmerksamkeitsblock mit relativen Positionseinbettungen.

Quellcode in ultralytics/models/sam/modules/encoders.py
class Attention(nn.Module):
    """Multi-head Attention block with relative position embeddings."""

    def __init__(
        self,
        dim: int,
        num_heads: int = 8,
        qkv_bias: bool = True,
        use_rel_pos: bool = False,
        rel_pos_zero_init: bool = True,
        input_size: Optional[Tuple[int, int]] = None,
    ) -> None:
        """
        Initialize Attention module.

        Args:
            dim (int): Number of input channels.
            num_heads (int): Number of attention heads.
            qkv_bias (bool):  If True, add a learnable bias to query, key, value.
            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
            input_size (tuple(int, int), None): Input resolution for calculating the relative
                positional parameter size.
        """
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = head_dim**-0.5

        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.proj = nn.Linear(dim, dim)

        self.use_rel_pos = use_rel_pos
        if self.use_rel_pos:
            assert input_size is not None, "Input size must be provided if using relative positional encoding."
            # Initialize relative positional embeddings
            self.rel_pos_h = nn.Parameter(torch.zeros(2 * input_size[0] - 1, head_dim))
            self.rel_pos_w = nn.Parameter(torch.zeros(2 * input_size[1] - 1, head_dim))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        """Applies the forward operation including attention, normalization, MLP, and indexing within window limits."""
        B, H, W, _ = x.shape
        # qkv with shape (3, B, nHead, H * W, C)
        qkv = self.qkv(x).reshape(B, H * W, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
        # q, k, v with shape (B * nHead, H * W, C)
        q, k, v = qkv.reshape(3, B * self.num_heads, H * W, -1).unbind(0)

        attn = (q * self.scale) @ k.transpose(-2, -1)

        if self.use_rel_pos:
            attn = add_decomposed_rel_pos(attn, q, self.rel_pos_h, self.rel_pos_w, (H, W), (H, W))

        attn = attn.softmax(dim=-1)
        x = (attn @ v).view(B, self.num_heads, H, W, -1).permute(0, 2, 3, 1, 4).reshape(B, H, W, -1)
        return self.proj(x)

__init__(dim, num_heads=8, qkv_bias=True, use_rel_pos=False, rel_pos_zero_init=True, input_size=None)

Initialisiere das Modul Aufmerksamkeit.

Parameter:

Name Typ Beschreibung Standard
dim int

Anzahl der EingangskanÀle.

erforderlich
num_heads int

Anzahl der Aufmerksamkeitsköpfe.

8
qkv_bias bool

Wenn True, fĂŒge einen lernbaren Bias zu query, key, value hinzu.

True
rel_pos_zero_init bool

Wenn True, werden die relativen Positionsparameter mit Null initialisiert.

True
input_size (tuple(int, int), None)

Eingabeauflösung fĂŒr die Berechnung der relativen GrĂ¶ĂŸe der Positionsparameter.

None
Quellcode in ultralytics/models/sam/modules/encoders.py
def __init__(
    self,
    dim: int,
    num_heads: int = 8,
    qkv_bias: bool = True,
    use_rel_pos: bool = False,
    rel_pos_zero_init: bool = True,
    input_size: Optional[Tuple[int, int]] = None,
) -> None:
    """
    Initialize Attention module.

    Args:
        dim (int): Number of input channels.
        num_heads (int): Number of attention heads.
        qkv_bias (bool):  If True, add a learnable bias to query, key, value.
        rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
        input_size (tuple(int, int), None): Input resolution for calculating the relative
            positional parameter size.
    """
    super().__init__()
    self.num_heads = num_heads
    head_dim = dim // num_heads
    self.scale = head_dim**-0.5

    self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
    self.proj = nn.Linear(dim, dim)

    self.use_rel_pos = use_rel_pos
    if self.use_rel_pos:
        assert input_size is not None, "Input size must be provided if using relative positional encoding."
        # Initialize relative positional embeddings
        self.rel_pos_h = nn.Parameter(torch.zeros(2 * input_size[0] - 1, head_dim))
        self.rel_pos_w = nn.Parameter(torch.zeros(2 * input_size[1] - 1, head_dim))

forward(x)

Wendet die VorwĂ€rtsoperationen einschließlich Aufmerksamkeit, Normalisierung, MLP und Indexierung innerhalb der Fenstergrenzen an.

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward(self, x: torch.Tensor) -> torch.Tensor:
    """Applies the forward operation including attention, normalization, MLP, and indexing within window limits."""
    B, H, W, _ = x.shape
    # qkv with shape (3, B, nHead, H * W, C)
    qkv = self.qkv(x).reshape(B, H * W, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
    # q, k, v with shape (B * nHead, H * W, C)
    q, k, v = qkv.reshape(3, B * self.num_heads, H * W, -1).unbind(0)

    attn = (q * self.scale) @ k.transpose(-2, -1)

    if self.use_rel_pos:
        attn = add_decomposed_rel_pos(attn, q, self.rel_pos_h, self.rel_pos_w, (H, W), (H, W))

    attn = attn.softmax(dim=-1)
    x = (attn @ v).view(B, self.num_heads, H, W, -1).permute(0, 2, 3, 1, 4).reshape(B, H, W, -1)
    return self.proj(x)



ultralytics.models.sam.modules.encoders.PatchEmbed

Basen: Module

Image to Patch Embedding.

Quellcode in ultralytics/models/sam/modules/encoders.py
class PatchEmbed(nn.Module):
    """Image to Patch Embedding."""

    def __init__(
        self,
        kernel_size: Tuple[int, int] = (16, 16),
        stride: Tuple[int, int] = (16, 16),
        padding: Tuple[int, int] = (0, 0),
        in_chans: int = 3,
        embed_dim: int = 768,
    ) -> None:
        """
        Initialize PatchEmbed module.

        Args:
            kernel_size (Tuple): kernel size of the projection layer.
            stride (Tuple): stride of the projection layer.
            padding (Tuple): padding size of the projection layer.
            in_chans (int): Number of input image channels.
            embed_dim (int): Patch embedding dimension.
        """
        super().__init__()

        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        """Computes patch embedding by applying convolution and transposing resulting tensor."""
        return self.proj(x).permute(0, 2, 3, 1)  # B C H W -> B H W C

__init__(kernel_size=(16, 16), stride=(16, 16), padding=(0, 0), in_chans=3, embed_dim=768)

Initialisiere das PatchEmbed-Modul.

Parameter:

Name Typ Beschreibung Standard
kernel_size Tuple

KernelgrĂ¶ĂŸe der Projektionsschicht.

(16, 16)
stride Tuple

schritt der Projektionsschicht.

(16, 16)
padding Tuple

AuffĂŒllgrĂ¶ĂŸe der Projektionsschicht.

(0, 0)
in_chans int

Anzahl der EingangsbildkanÀle.

3
embed_dim int

Patch-Einbettungsdimension.

768
Quellcode in ultralytics/models/sam/modules/encoders.py
def __init__(
    self,
    kernel_size: Tuple[int, int] = (16, 16),
    stride: Tuple[int, int] = (16, 16),
    padding: Tuple[int, int] = (0, 0),
    in_chans: int = 3,
    embed_dim: int = 768,
) -> None:
    """
    Initialize PatchEmbed module.

    Args:
        kernel_size (Tuple): kernel size of the projection layer.
        stride (Tuple): stride of the projection layer.
        padding (Tuple): padding size of the projection layer.
        in_chans (int): Number of input image channels.
        embed_dim (int): Patch embedding dimension.
    """
    super().__init__()

    self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding)

forward(x)

Berechnet die Patch-Einbettung durch Faltung und Transponierung des Ergebnisses tensor.

Quellcode in ultralytics/models/sam/modules/encoders.py
def forward(self, x: torch.Tensor) -> torch.Tensor:
    """Computes patch embedding by applying convolution and transposing resulting tensor."""
    return self.proj(x).permute(0, 2, 3, 1)  # B C H W -> B H W C



ultralytics.models.sam.modules.encoders.window_partition(x, window_size)

Unterteilung in nicht ĂŒberlappende Fenster mit AuffĂŒllung, falls nötig. Args: x (tensor): Eingabe-Token mit [B, H, W, C]. window_size (int): FenstergrĂ¶ĂŸe.

Retouren:

Name Typ Beschreibung
windows Tensor

Fenster nach Partitionierung mit [B * num_windows, window_size, window_size, C].

(Hp, Wp)

gepolsterte Höhe und Breite vor der Teilung

Quellcode in ultralytics/models/sam/modules/encoders.py
def window_partition(x: torch.Tensor, window_size: int) -> Tuple[torch.Tensor, Tuple[int, int]]:
    """
    Partition into non-overlapping windows with padding if needed.
    Args:
        x (tensor): input tokens with [B, H, W, C].
        window_size (int): window size.

    Returns:
        windows: windows after partition with [B * num_windows, window_size, window_size, C].
        (Hp, Wp): padded height and width before partition
    """
    B, H, W, C = x.shape

    pad_h = (window_size - H % window_size) % window_size
    pad_w = (window_size - W % window_size) % window_size
    if pad_h > 0 or pad_w > 0:
        x = F.pad(x, (0, 0, 0, pad_w, 0, pad_h))
    Hp, Wp = H + pad_h, W + pad_w

    x = x.view(B, Hp // window_size, window_size, Wp // window_size, window_size, C)
    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
    return windows, (Hp, Wp)



ultralytics.models.sam.modules.encoders.window_unpartition(windows, window_size, pad_hw, hw)

Fensteraufteilung in Originalsequenzen und Entfernen von FĂŒllmaterial.

Parameter:

Name Typ Beschreibung Standard
windows tensor

Eingabe-Token mit [B * num_windows, window_size, window_size, C].

erforderlich
window_size int

FenstergrĂ¶ĂŸe.

erforderlich
pad_hw Tuple

gepolsterte Höhe und Breite (Hp, Wp).

erforderlich
hw Tuple

ursprĂŒngliche Höhe und Breite (H, W) vor dem AuffĂŒllen.

erforderlich

Retouren:

Name Typ Beschreibung
x Tensor

unpartitionierte Sequenzen mit [B, H, W, C].

Quellcode in ultralytics/models/sam/modules/encoders.py
def window_unpartition(
    windows: torch.Tensor, window_size: int, pad_hw: Tuple[int, int], hw: Tuple[int, int]
) -> torch.Tensor:
    """
    Window unpartition into original sequences and removing padding.

    Args:
        windows (tensor): input tokens with [B * num_windows, window_size, window_size, C].
        window_size (int): window size.
        pad_hw (Tuple): padded height and width (Hp, Wp).
        hw (Tuple): original height and width (H, W) before padding.

    Returns:
        x: unpartitioned sequences with [B, H, W, C].
    """
    Hp, Wp = pad_hw
    H, W = hw
    B = windows.shape[0] // (Hp * Wp // window_size // window_size)
    x = windows.view(B, Hp // window_size, Wp // window_size, window_size, window_size, -1)
    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, Hp, Wp, -1)

    if Hp > H or Wp > W:
        x = x[:, :H, :W, :].contiguous()
    return x



ultralytics.models.sam.modules.encoders.get_rel_pos(q_size, k_size, rel_pos)

Erhalte relative Positionseinbettungen entsprechend den relativen Positionen von Abfrage und SchlĂŒsselgrĂ¶ĂŸen.

Parameter:

Name Typ Beschreibung Standard
q_size int

GrĂ¶ĂŸe der Abfrage q.

erforderlich
k_size int

GrĂ¶ĂŸe des SchlĂŒssels k.

erforderlich
rel_pos Tensor

relative Positionseinbettungen (L, C).

erforderlich

Retouren:

Typ Beschreibung
Tensor

Extrahierte Positionseinbettungen nach relativen Positionen.

Quellcode in ultralytics/models/sam/modules/encoders.py
def get_rel_pos(q_size: int, k_size: int, rel_pos: torch.Tensor) -> torch.Tensor:
    """
    Get relative positional embeddings according to the relative positions of query and key sizes.

    Args:
        q_size (int): size of query q.
        k_size (int): size of key k.
        rel_pos (Tensor): relative position embeddings (L, C).

    Returns:
        Extracted positional embeddings according to relative positions.
    """
    max_rel_dist = int(2 * max(q_size, k_size) - 1)
    # Interpolate rel pos if needed.
    if rel_pos.shape[0] != max_rel_dist:
        # Interpolate rel pos.
        rel_pos_resized = F.interpolate(
            rel_pos.reshape(1, rel_pos.shape[0], -1).permute(0, 2, 1),
            size=max_rel_dist,
            mode="linear",
        )
        rel_pos_resized = rel_pos_resized.reshape(-1, max_rel_dist).permute(1, 0)
    else:
        rel_pos_resized = rel_pos

    # Scale the coords with short length if shapes for q and k are different.
    q_coords = torch.arange(q_size)[:, None] * max(k_size / q_size, 1.0)
    k_coords = torch.arange(k_size)[None, :] * max(q_size / k_size, 1.0)
    relative_coords = (q_coords - k_coords) + (k_size - 1) * max(q_size / k_size, 1.0)

    return rel_pos_resized[relative_coords.long()]



ultralytics.models.sam.modules.encoders.add_decomposed_rel_pos(attn, q, rel_pos_h, rel_pos_w, q_size, k_size)

Berechne die dekomponierten Relativen Positionseinbettungen aus dem mvitv2-Papier unter https://github.com/facebookresearch/mvit/blob/main/mvit/models/attention.py.

Parameter:

Name Typ Beschreibung Standard
attn Tensor

Achtung Karte.

erforderlich
q Tensor

Abfrage q in der Aufmerksamkeitsschicht mit der Form (B, q_h * q_w, C).

erforderlich
rel_pos_h Tensor

relative Positionseinbettungen (Lh, C) fĂŒr die Höhenachse.

erforderlich
rel_pos_w Tensor

relative Positionseinbettungen (Lw, C) fĂŒr die Breitenachse.

erforderlich
q_size Tuple

rĂ€umliche SequenzgrĂ¶ĂŸe der Abfrage q mit (q_h, q_w).

erforderlich
k_size Tuple

rĂ€umliche SequenzgrĂ¶ĂŸe des SchlĂŒssels k mit (k_h, k_w).

erforderlich

Retouren:

Name Typ Beschreibung
attn Tensor

Aufmerksamkeitskarte mit zusÀtzlichen relativen Positionseinbettungen.

Quellcode in ultralytics/models/sam/modules/encoders.py
def add_decomposed_rel_pos(
    attn: torch.Tensor,
    q: torch.Tensor,
    rel_pos_h: torch.Tensor,
    rel_pos_w: torch.Tensor,
    q_size: Tuple[int, int],
    k_size: Tuple[int, int],
) -> torch.Tensor:
    """
    Calculate decomposed Relative Positional Embeddings from mvitv2 paper at
    https://github.com/facebookresearch/mvit/blob/main/mvit/models/attention.py.

    Args:
        attn (Tensor): attention map.
        q (Tensor): query q in the attention layer with shape (B, q_h * q_w, C).
        rel_pos_h (Tensor): relative position embeddings (Lh, C) for height axis.
        rel_pos_w (Tensor): relative position embeddings (Lw, C) for width axis.
        q_size (Tuple): spatial sequence size of query q with (q_h, q_w).
        k_size (Tuple): spatial sequence size of key k with (k_h, k_w).

    Returns:
        attn (Tensor): attention map with added relative positional embeddings.
    """
    q_h, q_w = q_size
    k_h, k_w = k_size
    Rh = get_rel_pos(q_h, k_h, rel_pos_h)
    Rw = get_rel_pos(q_w, k_w, rel_pos_w)

    B, _, dim = q.shape
    r_q = q.reshape(B, q_h, q_w, dim)
    rel_h = torch.einsum("bhwc,hkc->bhwk", r_q, Rh)
    rel_w = torch.einsum("bhwc,wkc->bhwk", r_q, Rw)

    attn = (attn.view(B, q_h, q_w, k_h, k_w) + rel_h[:, :, :, :, None] + rel_w[:, :, :, None, :]).view(
        B, q_h * q_w, k_h * k_w
    )

    return attn





Erstellt am 2023-11-12, Aktualisiert am 2023-11-25
Autoren: glenn-jocher (3), Laughing-q (1)