Page 353 - AI Ver 3.0 class 10_Flipbook
P. 353

First, we have the input layer which receives the input in several different formats provided by the programmer and
                 feeds it to the neural network. Minimal processing occurs in the input layer, as it simply passes the raw input data
                 forward. The output layer produces the final prediction or decision based on the learned patterns. The output at
                 each node is called its activation or node value.

                         What is Convolutional Neural Network (CNN)?


                 Convolutional Neural Network is a type of Artificial Neural Network and is made up of neurons that help in image
                 recognition and image processing. It uses a deep learning algorithm that takes an input image, processes it by
                 assigning learnable weights and biases to various aspects/objects in the image, enabling the network to identify
                 patterns and features helping the system differentiate one image from the other with maximum accuracy. CNNs
                 reduce the spatial dimensions (size) of the input through operations like pooling, while retaining the essential
                 features and give the predicted class probabilities for the input image. They are trained to identify and extract the
                 best features from the images.



                                                                                               CARGO SHIP
                                                                                               STEAM BOAT
                                                                                               CRUISE      Cargo Ship  17%
                                                                                                           Steam Boat  8%
                                                                                               FISHING BOAT  Cruise  75%
                                          INPUT  CONVOLUTION - ReLU  POOLING  CONVOLUTION - ReLU  POOLING  FLATTEN  FULLY   SOFTMAX
                                                                                       CONNECTED
                                                                                                          Output Probability
                                                            FEATURE LEARNING          CLASSIFICATION
                    Input Image
                                                      Image Processed by CNN
                 In the above diagram, an input image is provided, processed through a CNN, and a prediction is generated based
                 on the labels in the corresponding dataset.


                               Brainy Fact


                      Convolutional  neural  networks  (ConvNets)  were  first  introduced  in  the  1980s  by Yann  LeCun,  a  computer
                      science researcher. Its early version called LeNet (after LeCun), were used to recognise handwritten digits. It
                      found its use in postal services to read zip codes on envelopes and in banking/financial sectors to read digits
                      on cheques.



                         Layers of Convolutional Neural Network (CNN)


                 The different layers of a Convolutional Neural Network (CNN) are shown in the following figure:




                                                                                                           CARGO SHIP
                                                                                                           STEAM BOAT
                                                                                                           CRUISE




                                                                                                           FISHING BOAT

                           INPUT   CONVOLUTION - ReLU  POOLING  CONVOLUTION - ReLU  POOLING  FLATTEN  FULLY   SOFTMAX
                                                                                               CONNECTED
                                                      FEATURE LEARNING                        CLASSIFICATION
                                                                                    Computer Vision (Practical)  351
   348   349   350   351   352   353   354   355   356   357   358