Introduction: Image classification is the most crucial domain in which neural networks play a significant role....

Transcribed Image Text:

Introduction: Image classification is the most crucial domain in which neural networks play a significant role. The neural network accepts the input images and provides an output classification. In this project, develop an image classifier using basic neural networks with few different optimizers. Dataset: The MNIST digits dataset provided by Yann Le Cun, is a well-known dataset for classification. It includes images with 255 pixels 28 x 28 greyscale and labelled with a number from 0 to 9, denoting the appropriate digit. There are 60,000 training images and 10,000 testing datasets. 0000 Task 2: Import Dataset: 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 9 9 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 66 6 6 6 6 6 6 6 6 6 6 6 6 6 7777777777777777 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 MNIST Digit dataset Dataset can be import using keras API: https://keras.io/api/datasets/ Task 1: Import Libraries: a. Import required libraries (e.g., import tensorflow as tf) a. Load dataset using keras API. b. Dataset must be pre-processed before training the network, if you check image in the training, you will see pixel value range from 0 to 255, scale these values range from 0 to 1. [note it is important that both training and testing dataset pre- processed same way]. 7346 61 8 10 9 8 03127 02960 167 19 Verifying MNIST Digit data is in correct format Task 3: Build a Classifier using MLP a. The layer is the most fundamental component of a neural network. Data is put into layers, and they extract representations from it. Choose the number of stacking layers so that model representations should be useful for the given task. Task 4: Compile the Model a. A few more parameters are required before the model can be used for training. These are added at the build step of the model: Choose these parameters. a.1 Loss function a.2 Optimizer a.3 Metrics Task 5: Train and Test the model. a. Feed the training data to the built model. b. Ask model to make predictions about a test set. c. Verify that the predictions match the labels from the test labels. Prepare a report: (1000-1200 words) Your report should contain the following: 1) Introduction 2) Literature Review (In literature review provide comprehensive summary of previous work). 3) The methods applied for solving each task and reason for choosing. 4) Results: Take a screenshot of the results and attach it in document. 5) Challenges and problem during project 6) References: cite all the information used in this report. 7) Applies integrity system and ethical theories to evaluate all possible ethical issues in this scenario, and demonstrate your understanding of IT Professional ethics by proposing solutions for the ethical issues discussed. Introduction: Image classification is the most crucial domain in which neural networks play a significant role. The neural network accepts the input images and provides an output classification. In this project, develop an image classifier using basic neural networks with few different optimizers. Dataset: The MNIST digits dataset provided by Yann Le Cun, is a well-known dataset for classification. It includes images with 255 pixels 28 x 28 greyscale and labelled with a number from 0 to 9, denoting the appropriate digit. There are 60,000 training images and 10,000 testing datasets. 0000 Task 2: Import Dataset: 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 9 9 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 66 6 6 6 6 6 6 6 6 6 6 6 6 6 7777777777777777 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 MNIST Digit dataset Dataset can be import using keras API: https://keras.io/api/datasets/ Task 1: Import Libraries: a. Import required libraries (e.g., import tensorflow as tf) a. Load dataset using keras API. b. Dataset must be pre-processed before training the network, if you check image in the training, you will see pixel value range from 0 to 255, scale these values range from 0 to 1. [note it is important that both training and testing dataset pre- processed same way]. 7346 61 8 10 9 8 03127 02960 167 19 Verifying MNIST Digit data is in correct format Task 3: Build a Classifier using MLP a. The layer is the most fundamental component of a neural network. Data is put into layers, and they extract representations from it. Choose the number of stacking layers so that model representations should be useful for the given task. Task 4: Compile the Model a. A few more parameters are required before the model can be used for training. These are added at the build step of the model: Choose these parameters. a.1 Loss function a.2 Optimizer a.3 Metrics Task 5: Train and Test the model. a. Feed the training data to the built model. b. Ask model to make predictions about a test set. c. Verify that the predictions match the labels from the test labels. Prepare a report: (1000-1200 words) Your report should contain the following: 1) Introduction 2) Literature Review (In literature review provide comprehensive summary of previous work). 3) The methods applied for solving each task and reason for choosing. 4) Results: Take a screenshot of the results and attach it in document. 5) Challenges and problem during project 6) References: cite all the information used in this report. 7) Applies integrity system and ethical theories to evaluate all possible ethical issues in this scenario, and demonstrate your understanding of IT Professional ethics by proposing solutions for the ethical issues discussed.