Page 266 - Ai Book - 10

u Fully Connected Layer: The last layer in the
Convolution Neural Network is the Fully
Connected layer. The basic objective of fully
connected layer is to classify the images on
the basis of input. The result of convolution/
pooling process acts as an input for for fully
connected layer. In a fully connected layer, the
output of convolution/pooling is flattened into
a single vector of values, each representing a
probability that a certain feature belongs to a
label.
Simply put, fully connected layer is responsible to predict the best label for an image on the basis of the
result of convolution or pooling layer.

K Keyey TTermserms

u Computer Vision
Computer vision teaches machines to interpret and understand the visual world, processing information
from images. At allows computers to analyse and comprehend visual data, such as recognizing objects or

patterns.

u Instance Segmentation
Predicting individual object instances and creating pixel masks to distinguish them. Separating and outlining
distinct people in a crowded image.

u Pixels
The smallest units of a digital image, representing individual points of color. Each dot in a digital photograph
is a pixel, collectively forming the image.

u Resolution
The total pixel count in a digital image, influencing the image’s clarity and detail. A higher-resolution image
contains more pixels, providing finer detail.

u Convolution Network
Applying convolution as a technique for image processing, involving convolution operators and kernels.
Blurring or sharpening an image using convolution to enhance or modify its appearance.

u Neural Network
A system of algorithms modeled after the human brain, comprising layers with weights and biases for
pattern recognition. Training a neural network to recognize patterns in data, such as distinguishing between
handwritten digits.

u Convolutional Neural Network (CNN)
A specialized neural network designed for image-related tasks, featuring convolutional layers for feature
extraction. Using a CNN for image recognition applications such as identifying objects in photos or medical
image classification.

140
140