MobileNet

ConvNet Image Classification

class lucid.models.MobileNet(width_multiplier: float, num_classes: int = 1000)

Overview

The MobileNet class implements the MobileNet-v1 architecture, which introduces depthwise separable convolutions to reduce computational cost while maintaining accuracy. This architecture is ideal for mobile and embedded vision applications.

MobileNet architecture

Class Signature

class MobileNet(nn.Module):
    def __init__(self, width_multiplier: float, num_classes: int = 1000) -> None

Parameters

  • width_multiplier (float): Adjusts the width of the network by scaling the number of channels in each layer. A higher value increases the capacity of the model, while a lower value reduces computational cost.

  • num_classes (int, optional): Number of output classes for the classification task. Default is 1000, commonly used for ImageNet.

Tip

  • Adjust the width_multiplier and num_classes parameters to suit specific datasets or computational constraints.