Inception

ConvNet Image Classification

class lucid.models.Inception(num_classes: int, use_aux: bool | None = True)

Overview

The Inception base class provides a flexible implementation for defining various versions of the Inception architecture, including Inception v1, v3, and v4.

It facilitates the configuration of the feature extraction and classification components through arguments, making it adaptable for different versions of the Inception series.

Inception-v4 architecture

Class Signature

class Inception(nn.Module):
    def __init__(self, num_classes: int, use_aux: bool = True) -> None

Parameters

  • num_classes (int) The number of output classes for the final classification layer.

  • use_aux (bool, optional) Whether to include auxiliary classifiers. Auxiliary classifiers are additional branches used during training to assist optimization. Default is True.