Skip to content

Computer Vision

tensoris.backend.components.blocks.cv

Computer Vision Network Blocks.

Classes

ResidualBlock

Bases: Module

Residual Skip Connection Block for ResNet architectures.

References

He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep Residual Learning for Image Recognition. IEEE CVPR 2016. arXiv: https://arxiv.org/abs/1512.03385

Source code in src/tensoris/backend/components/blocks/cv.py 
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
class ResidualBlock(nn.Module):
    """Residual Skip Connection Block for ResNet architectures.

    References:
        He, K., Zhang, X., Ren, S., & Sun, J. (2016).
        Deep Residual Learning for Image Recognition. IEEE CVPR 2016.
        arXiv: https://arxiv.org/abs/1512.03385
    """

    def __init__(self, channels: int) -> None:
        """Initialize ResidualBlock.

        Args:
            channels: Feature map channel dimension.
        """
        super().__init__()
        self.conv1 = nn.Conv2d(channels, channels, kernel_size=3, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(channels)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(channels, channels, kernel_size=3, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(channels)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        """Execute ResidualBlock forward pass with identity shortcut."""
        residual = x
        out = self.relu(self.bn1(self.conv1(x)))
        out = self.bn2(self.conv2(out))
        out += residual
        return self.relu(out)
Methods:
__init__ 
__init__(channels)

Initialize ResidualBlock.

Parameters:

Name Type Description Default
channels int

Feature map channel dimension.

 required
Source code in src/tensoris/backend/components/blocks/cv.py 
16
17
18
19
20
21
22
23
24
25
26
27
def __init__(self, channels: int) -> None:
    """Initialize ResidualBlock.

    Args:
        channels: Feature map channel dimension.
    """
    super().__init__()
    self.conv1 = nn.Conv2d(channels, channels, kernel_size=3, padding=1, bias=False)
    self.bn1 = nn.BatchNorm2d(channels)
    self.relu = nn.ReLU(inplace=True)
    self.conv2 = nn.Conv2d(channels, channels, kernel_size=3, padding=1, bias=False)
    self.bn2 = nn.BatchNorm2d(channels)
forward 
forward(x)

Execute ResidualBlock forward pass with identity shortcut.

Source code in src/tensoris/backend/components/blocks/cv.py 
29
30
31
32
33
34
35
def forward(self, x: torch.Tensor) -> torch.Tensor:
    """Execute ResidualBlock forward pass with identity shortcut."""
    residual = x
    out = self.relu(self.bn1(self.conv1(x)))
    out = self.bn2(self.conv2(out))
    out += residual
    return self.relu(out)