Page 222 - Ai_C10_Flipbook
P. 222
The pooling layer is important in the CNN as it performs a series of tasks which are as follows:
• Pooling reduces the spatial dimensions (height and width) of the feature map, which helps reduce the number
of parameters and computation in the network. This makes the model faster and more efficient.
• Even though the size is reduced, the most important features (like edges and patterns) are preserved, ensuring
that the network can still make accurate predictions based on the crucial information.
• Pooling helps make the model more robust to variations in the input image (e.g., slight changes in position,
rotation, or scale of objects in the image). Since pooling looks at local regions, it makes the network less
sensitive to small transformations.
• By reducing the spatial dimensions, pooling helps lower the risk of overfitting by simplifying the model,
ensuring it generalises well to new, unseen data.
Max
Pooling
Sum
A small difference in the input image will create a very similar pooled image.
Fully Connected Layer
This is the last and the final layer of the Convolutional Neural Network. After the features of the input image
are extracted by the convolution layers and downsampled by the pooling layers, their output is a 3-dimensional
matrix which is flattened into a vector of values. These values of the single vector represent a specific feature of a
specific label and are redirected to fully connected layers to predict the final outputs of the network. This helps in
classifying an image into a specific label based on the probability of the input being in a specific class. For example,
the input image with a beak belongs to the category of a bird.
bird p bird
sunset p sunset
dog p dog
... p
cat cat
Convolution + Nonlinearity Max pooling Vec ...
Convolution + Pooling layers Fully connected layers Nx binary classification
Brainy Fact
The first image recognition CNN-AlexNet named after(creator-Alex Krizhevsky), won the 2012 ImageNet
Computer Vision contest with 85% accuracy.
220 Artificial Intelligence Play (Ver 1.0)-X

