Neural Nets for Object Recognition: CIFAR-10 Dataset

Building Neural Networks for Object Recognition with CIFAR-10 Dataset

On the completion of this assignment you will be able to:

Critically appraise a comprehensive/detailed understanding of the computational intelligence domain.
Design and develop computational intelligence software artefacts.
Critique and contextualise emerging research in the area of computational intelligence.

The Cardiff Met EDGE supports students in graduating with the knowledge, skills, and attributes that allow them to contribute positively and effectively to the communities in which they live and work.

This module assessment provides opportunities for students to demonstrate development of the following EDGE Competencies:

ETHICAL	Students will be required to consider Ethical implication of their analysis and follow the necessary ethical approval processes while addressing problems associated with the Object Recognition assessment.
DIGITAL	Students will be required to demonstrate digital skills in the manipulation of image data and analysis for their project.
GLOBAL	Students will demonstrate an awareness of the global context of Object Recognition and apply this to their assessment
ENTREPRENEURIAL	Students will also demonstrate their developed entrepreneurial through working under their own initiative, formulating and presenting recommendations in order to solve an authentic and complex problem associated with the module.

Nowadays, Computational Intelligence (CI) artefacts are powering various industries and sectors globally. Many CI products, like self-driving cars, for example, utilize object detection. The ability to build intelligent products that can detect objects is a valuable skill. Some other interesting applications of object detection include: Face detection (in the new iPhone), object tracking, people counting, pedestrian detection, video surveillance etc.

In this task, you are required to build two Neural Networks of different structures using the CIFAR-10- Object Recognition image dataset.

“CIFAR-10 hosted in Kaggle, and directly in keras-datasets, is an established computer-vision dataset used for object recognition. It is a subset of the 80 million tiny images dataset, and consists of 60,000 32x32 colour images containing one of 10 object classes, with 6000 images per class. It was collected by Alex Krizhevsky, Vinod Nair, and Geoffrey Hinton.”

Your overall task is to build four (4) neural networks with different structures and evaluate their performances. You are expected to create a validation set from the training set. Hence you would have:

* A Training set

* A Validation set

* A Testing set

You are expected to create a report which should contain the following:

For each of your four (4) models, you will need to clearly show in a table, the differences in structure, hyperparameters and evaluation results. Furthermore:

Give a brief introduction about your data.
Describe how you created your validation set from your training set and give details of the data (metadata). Also, discuss the rationale for having a validation set.
Use a Table to Describe the structure, hyperparameters and evaluation results for each of your models.
For your Best performing model, Do the following:
1. Discuss the activation function used
2. Discuss the loss function used
3. Give an account of the number of epochs used.
4. Discuss why you feel this particular model outperformed all the other models while reflecting on the results it outputted.
5. For the Evaluation, You are also expected to use appropriate plots/ graphs and a short description (of your own interpretation) of the results.
6. Evaluate only the best model on your test dataset.