EfficientNet_V2

ConvNet Image Classification

class lucid.models.EfficientNet_V2(block_cfg: list, num_classes: int = 1000, dropout: float = 0.2, drop_path_rate: float = 0.2)

EfficientNet_V2 builds on the EfficientNet architecture, which employs a compound scaling method to balance depth, width, and resolution for optimal performance. The V2 variant introduces further improvements such as training with larger batch sizes, using higher-resolution images, and advanced regularization techniques like stochastic depth and progressive learning.

EfficientNet-v2 architecture

Class Signature

class EfficientNet_V2(nn.Module):
    def __init__(
        self,
        block_cfg: list,
        num_classes: int = 1000,
        dropout: float = 0.2,
        drop_path_rate: float = 0.2,
    ) -> None

Parameters

  • block_cfg (list): A list defining the structure and parameters of the building blocks in the network. Each entry specifies the configuration of a block, such as number of filters, stride, etc.

  • num_classes (int, optional): The number of output classes for classification. Default is 1000.

  • dropout (float, optional): The dropout rate applied to the final fully connected layer. Default is 0.2.

  • drop_path_rate (float, optional): The rate for stochastic depth regularization. Default is 0.2.

Warning

Ensure the block_cfg is well-defined to avoid shape mismatches or runtime errors during the forward pass.