ConvNet Architecture in CNN

Back to: Data Science Tutorials

ConvNet Architecture in CNN

In this article, I am going to discuss ConvNet Architecture in CNN. Please read our previous article where we discussed SoftMax vs Cross Entropy in CNN.

ConvNet Architecture in CNN

Convolutional Neural Networks (ConvNets/CNNs) is a form of neural network that is built primarily for processing data having a grid-like layout, such as images. CNN’s are very helpful for picture classification and identification.

A ConvNet’s architecture is intended to take use of the 2D structure of an input picture. A ConvNet is made up of layers such as convolutional layers, pooling layers, and fully-connected layers.

Convolutional Layers

The basic building block of a ConvNet is convolutional layers. They are in charge of extracting features from input data. A convolutional layer applies to the input data a collection of learnable filters, each of which is responsible for recognizing a distinct characteristic in the data.

Each filter is tiny, covering only a small portion of the incoming data. The filters are applied in a sliding window way to the input data, with the filter being convolved with a tiny section of the input data known as the receptive field.

As the filter moves through the input data, it generates a feature map that indicates the existence of the feature recognized by the filter in the data. Because the filter only covers a tiny section of the input data at a time, the feature map is smaller than the input data.

Multiple filters can be applied to the input data, resulting in a separate feature map for each. A convolutional layer’s output is therefore a series of feature maps that reflect the existence of various characteristics in the input data.

Pooling Layers

Pooling layers are used to reduce the spatial dimensions of the feature maps produced by convolutional layers. This reduces the amount of data, making ConvNet quicker and more efficient. Pooling may be classified into two types: maximum pooling and average pooling.

Max pooling extracts the largest value from a feature map area, whereas average pooling extracts the average value. Pooling is generally applied individually to each feature map, with the pooling window sliding over the feature map in the same way as the convolutional filters slide over the input data.

Fully-Connected Layers

Fully-connected layers are used to classify the features extracted by the convolutional layers. They take in the flattened feature maps produced by the previous layers and output a prediction.

In a fully-connected layer, each neuron in the layer is connected to every neuron in the previous layer. The neurons in a fully-connected layer learn a weighted sum of the inputs, which is then passed through an activation function to produce the output.

In a ConvNet, the fully-connected layers are typically placed at the end of the network, after the convolutional and pooling layers. They take in the flattened feature maps and use them to classify the input data.

In addition to these layers, a ConvNet may also include other layers such as normalization layers (e.g. batch normalization) and activation layers (e.g. ReLU).

In the next article, I am going to discuss Overfitting and Regularization in CNN. Here, in this article, I try to explain ConvNet Architecture in CNN. I hope you enjoy this ConvNet Architecture in the CNN article. Please post your feedback, suggestions, and questions about this article.