COMP338 – Computer Vision – Assignment 2

• This assignment is worth 15% of the total mark for COMP338
• Students will do the assignment individually.

Submission Instructions

• Send all solutions as a single PDF document containing your answers, results, and discussion of the results. Attach the source code for the programming problems as separate files.
• Each student will make a single submission to the Canvas system.
• The deadline for this assignment 09/12/2022, 5:00pm
• Penalties for late submission apply in accordance with departmental policy as set out in the student handbook, which can be found at http://intranet.csc.liv.ac.uk/student/msc-handbook.pdf and the University Code of Practice on Assessment, found at https://www.liverpool.ac.uk/media/livacuk/tqsd/code-of-practice-on-assessment/code_of_practice_on_assessment.pdf

Image Classification with CNN

In this project, we will do image classification using the Fashion MNIST dataset. The lab “COMP338_Lab_08_Fashion_MNIST_Classification.ipynb” on Canvas shows the example source code for this assignment.

Tasks:

1. (30 marks) Design a deep neural network for image classification.
2. (30 marks) Train and test your network on Fashion MNIST dataset.
3. (40 marks) Write a report to clearly explain your network, the intuition behind your design, and discussion of your results.

Rules:

• You can refer to any papers and reuse any source code. However, you should clearly cite the references in your report.
• Use free Google Colab account (https://colab.research.google.com/) for training. The maximum training time on a free Google Colab account is 12 hours.

Our solution will be evaluated by:

• The robustness of your network design (30%).
• The accuracy of your trained model, compared with other students (30%).
• The completeness of your report (40%).