minerva.models.nets.dcnn

Classes

DCNN	A Deep Convolutional Neural Network (DCNN) for Human Activity Recognition (HAR)

Module Contents

class minerva.models.nets.dcnn.DCNN(num_classes=6)

Bases: torch.nn.Module

A Deep Convolutional Neural Network (DCNN) for Human Activity Recognition (HAR) using wearable sensors, based on the methodology presented in the paper “Human Activity Recognition using Wearable Sensors by Deep Convolutional Neural Networks.”

This DCNN architecture is designed to process 2D activity images generated from wearable sensor signals and classify them into different activity categories.

Initializes the DCNN with the given number of output classes.

Parameters:

num_classes: int, optional

The number of activity classes. Defaults to 6.

conv1

conv2

fc1

fc2

forward(x)

Defines the forward pass of the DCNN.

This method processes the input 2D activity images through a series of convolutional and pooling layers, followed by fully connected layers to produce the final class probabilities.

Parameters:

x: torch.Tensor

The input tensor representing the batch of 2D activity images. Shape should be (batch_size, 1, height, width).

Returns:

torch.Tensor:

The output tensor containing the class probabilities for each input. Shape will be (batch_size, num_classes).

pool1

pool2