Vai al contenuto

Riferimento per ultralytics/models/sam/modules/encoders.py

Nota

Questo file è disponibile su https://github.com/ultralytics/ ultralytics/blob/main/ ultralytics/models/ sam/modules/encoders .py. Se riscontri un problema, contribuisci a risolverlo inviando una Pull Request 🛠️. Grazie 🙏!



ultralytics.models.sam.modules.encoders.ImageEncoderViT

Basi: Module

Un codificatore di immagini che utilizza l'architettura Vision Transformer (ViT) per codificare un'immagine in uno spazio latente compatto. Il codificatore prende un'immagine, la divide in patch e le elabora attraverso una serie di blocchi trasformatori. Le patch codificate vengono poi elaborate attraverso un collo per generare la rappresentazione finale codificata.

Questa classe e le sue funzioni di supporto di seguito riportate sono state leggermente adattate dalla struttura portante di ViTDet disponibile all'indirizzo https://github.com/facebookresearch/detectron2/blob/main/detectron2/modeling/backbone/vit.py.

Attributi:

Nome Tipo Descrizione
img_size int

Dimensione delle immagini in ingresso, che si presume sia quadrata.

patch_embed PatchEmbed

Modulo per l'incorporazione di patch.

pos_embed Parameter

Incorporamento posizionale assoluto per le patch.

blocks ModuleList

Elenco di blocchi trasformatori per l'elaborazione di patch embeddings.

neck Sequential

Modulo Neck per elaborare ulteriormente l'output.

Codice sorgente 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=())

Parametri:

Nome Tipo Descrizione Predefinito
img_size int

Dimensione dell'immagine in ingresso.

1024
patch_size int

Dimensioni della patch.

16
in_chans int

Numero di canali immagine in ingresso.

3
embed_dim int

Dimensione di incorporazione della patch.

768
depth int

Profondità del ViT.

12
num_heads int

Numero di teste di attenzione in ogni blocco ViT.

12
mlp_ratio float

Rapporto tra la dimensione nascosta del mlp e la dimensione dell'incorporazione.

4.0
qkv_bias bool

Se Vero, aggiunge un pregiudizio imparabile a query, chiave, valore.

True
norm_layer Module

Livello di normalizzazione.

LayerNorm
act_layer Module

Strato di attivazione.

GELU
use_abs_pos bool

Se Vero, utilizza le incorporazioni posizionali assolute.

True
use_rel_pos bool

Se Vero, aggiunge le incorporazioni posizionali relative alla mappa dell'attenzione.

False
rel_pos_zero_init bool

Se Vero, azzera l'inizializzazione dei parametri posizionali relativi.

True
window_size int

Dimensione della finestra per i blocchi di attenzione della finestra.

0
global_attn_indexes list

Indici per i blocchi che utilizzano l'attenzione globale.

()
Codice sorgente 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)

Elabora l'input attraverso il patch embedding, applica l'embedding posizionale se presente e passa attraverso i blocchi e il collo. e collo.

Codice sorgente 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

Basi: Module

Codifica diversi tipi di messaggi, tra cui punti, caselle e maschere, da inserire nel decodificatore di maschere di SAM. Il codificatore produce sia incorporazioni rade che dense per i messaggi in ingresso.

Attributi:

Nome Tipo Descrizione
embed_dim int

Dimensione delle incorporazioni.

input_image_size Tuple[int, int]

Dimensioni dell'immagine di ingresso (H, W).

image_embedding_size Tuple[int, int]

Dimensione spaziale dell'immagine incorporata (H, W).

pe_layer PositionEmbeddingRandom

Modulo per l'incorporazione di posizioni casuali.

num_point_embeddings int

Numero di incorporazioni di punti per diversi tipi di punti.

point_embeddings ModuleList

Elenco di incorporazioni di punti.

not_a_point_embed Embedding

Incorporamento per i punti che non fanno parte di alcuna etichetta.

mask_input_size Tuple[int, int]

Dimensione della maschera di ingresso.

mask_downscaling Sequential

Rete neurale per il downscaling della maschera.

no_mask_embed Embedding

Incorporamento per i casi in cui non viene fornita alcuna maschera.

Codice sorgente 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)

Codifica le richieste da inserire nel decodificatore di maschere di SAM.

Parametri:

Nome Tipo Descrizione Predefinito
embed_dim int

La dimensione di incorporazione dei suggerimenti

richiesto
image_embedding_size tuple(int, int

La dimensione spaziale dell'immagine dell'immagine incorporata, come (H, W).

richiesto
input_image_size int

La dimensione imbottita dell'immagine in ingresso al codificatore di immagini, come (H, W).

richiesto
mask_in_chans int

Il numero di canali nascosti utilizzati per codificare le maschere di ingresso.

richiesto
activation Module

L'attivazione da utilizzare per la codifica delle delle maschere di input.

GELU
Codice sorgente 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)

Incorpora diversi tipi di richieste, restituendo sia incorporazioni rade che dense.

Parametri:

Nome Tipo Descrizione Predefinito
points (tuple(Tensor, Tensor), None)

le coordinate dei punti e le etichette da incorporare.

richiesto
boxes (Tensor, None)

scatole da incorporare

richiesto
masks (Tensor, None)

maschere da incorporare

richiesto

Restituzione:

Tipo Descrizione
Tensor

torch.Tensor: incorporazioni rade per i punti e le caselle, con forma BxNx(embed_dim), dove N è determinato dal numero di punti e caselle in ingresso. dal numero di punti e caselle in ingresso.

Tensor

torch.Tensor: incorporazioni dense per le maschere, nella forma Bx(embed_dim)x(embed_H)x(embed_W)

Codice sorgente 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()

Restituisce la codifica posizionale utilizzata per codificare le indicazioni dei punti, applicata a un insieme denso di punti della forma della codifica dell'immagine. codifica dell'immagine.

Restituzione:

Tipo Descrizione
Tensor

torch.Tensor: Codifica posizionale con forma 1x(embed_dim)x(embedding_h)x(embedding_w)

Codice sorgente 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

Basi: Module

Codifica posizionale con frequenze spaziali casuali.

Codice sorgente 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)

Inizializza un'incorporazione di posizione utilizzando frequenze spaziali casuali.

Codice sorgente 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)

Genera la codifica posizionale per una griglia della dimensione specificata.

Codice sorgente 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)

Codifica in modo posizionale i punti che non sono normalizzati a [0,1].

Codice sorgente 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

Basi: Module

Blocchi trasformatori con supporto di attenzione alle finestre e blocchi di propagazione residua.

Codice sorgente 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)

Parametri:

Nome Tipo Descrizione Predefinito
dim int

Numero di canali di ingresso.

richiesto
num_heads int

Numero di teste di attenzione in ogni blocco ViT.

richiesto
mlp_ratio float

Rapporto tra la dimensione nascosta del mlp e la dimensione dell'incorporazione.

4.0
qkv_bias bool

Se Vero, aggiunge un pregiudizio imparabile a query, chiave, valore.

True
norm_layer Module

Livello di normalizzazione.

LayerNorm
act_layer Module

Strato di attivazione.

GELU
use_rel_pos bool

Se Vero, aggiunge le incorporazioni posizionali relative alla mappa dell'attenzione.

False
rel_pos_zero_init bool

Se Vero, azzera l'inizializzazione dei parametri posizionali relativi.

True
window_size int

Dimensione della finestra per i blocchi di attenzione della finestra. Se è uguale a 0, allora utilizza l'attenzione globale.

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

Risoluzione di ingresso per il calcolo della dimensione relativa del parametro dimensione del parametro posizionale.

None
Codice sorgente 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)

Esegue un passaggio in avanti attraverso il blocco trasformatore con attenzione alle finestre e finestre non sovrapposte.

Codice sorgente 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

Basi: Module

Blocco di attenzione multi-testa con incorporazioni di posizione relativa.

Codice sorgente 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)

Inizializza il modulo Attention.

Parametri:

Nome Tipo Descrizione Predefinito
dim int

Numero di canali di ingresso.

richiesto
num_heads int

Numero di teste di attenzione.

8
qkv_bias bool

Se Vero, aggiunge un pregiudizio imparabile a query, chiave, valore.

True
rel_pos_zero_init bool

Se Vero, azzera l'inizializzazione dei parametri posizionali relativi.

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

Risoluzione di ingresso per il calcolo della dimensione relativa del parametro dimensione del parametro posizionale.

None
Codice sorgente 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)

Applica l'operazione di avanzamento, compresa l'attenzione, la normalizzazione, la MLP e l'indicizzazione entro i limiti della finestra.

Codice sorgente 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

Basi: Module

Inclusione di immagini in patch.

Codice sorgente 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)

Inizializza il modulo PatchEmbed.

Parametri:

Nome Tipo Descrizione Predefinito
kernel_size Tuple

dimensione del kernel del livello di proiezione.

(16, 16)
stride Tuple

del livello di proiezione.

(16, 16)
padding Tuple

dimensione del padding del livello di proiezione.

(0, 0)
in_chans int

Numero di canali immagine in ingresso.

3
embed_dim int

Dimensione di incorporazione della patch.

768
Codice sorgente 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)

Calcola il patch embedding applicando la convoluzione e trasponendo il risultato tensor.

Codice sorgente 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)

Suddivide le finestre in finestre non sovrapposte con un padding, se necessario. Args: x (tensor): token di input con [B, H, W, C]. window_size (int): dimensione della finestra.

Restituzione:

Nome Tipo Descrizione
windows Tensor

finestre dopo la partizione con [B * num_windows, window_size, window_size, C].

(Hp, Wp)

altezza e larghezza imbottite prima della suddivisione

Codice sorgente 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)

Dispartizione della finestra in sequenze originali e rimozione del padding.

Parametri:

Nome Tipo Descrizione Predefinito
windows tensor

token di input con [B * num_windows, window_size, window_size, C].

richiesto
window_size int

dimensione della finestra.

richiesto
pad_hw Tuple

altezza e larghezza imbottite (Hp, Wp).

richiesto
hw Tuple

altezza e larghezza originali (H, W) prima del padding.

richiesto

Restituzione:

Nome Tipo Descrizione
x Tensor

sequenze non partizionate con [B, H, W, C].

Codice sorgente 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)

Ottiene le incorporazioni posizionali relative in base alle posizioni relative della query e delle dimensioni della chiave.

Parametri:

Nome Tipo Descrizione Predefinito
q_size int

dimensione della query q.

richiesto
k_size int

dimensione della chiave k.

richiesto
rel_pos Tensor

le incorporazioni di posizione relativa (L, C).

richiesto

Restituzione:

Tipo Descrizione
Tensor

Estrazione delle incorporazioni posizionali in base alle posizioni relative.

Codice sorgente 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)

Calcola le Matrici Posizionali Relative decomposte dal documento mvitv2 all'indirizzo https://github.com/facebookresearch/mvit/blob/main/mvit/models/attention.py.

Parametri:

Nome Tipo Descrizione Predefinito
attn Tensor

mappa di attenzione.

richiesto
q Tensor

query q nel livello di attenzione con forma (B, q_h * q_w, C).

richiesto
rel_pos_h Tensor

Incorporamenti di posizione relativa (Lh, C) per l'asse dell'altezza.

richiesto
rel_pos_w Tensor

Incorporamenti di posizione relativa (Lw, C) per l'asse della larghezza.

richiesto
q_size Tuple

dimensione della sequenza spaziale della query q con (q_h, q_w).

richiesto
k_size Tuple

dimensione della sequenza spaziale della chiave k con (k_h, k_w).

richiesto

Restituzione:

Nome Tipo Descrizione
attn Tensor

mappa dell'attenzione con l'aggiunta di incorporazioni posizionali relative.

Codice sorgente 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





Created 2023-11-12, Updated 2024-06-02
Authors: glenn-jocher (5), Burhan-Q (1), Laughing-q (1)