The course will describe the theory and practice of deep Neural Networks, otherwise known as Deep Learning, with a particular emphasis on their use in Image Processing and Computer Vision.

Learning Outcomes

Upon successful completion, students will have the knowledge and skills to:

1. Describe and apply the theory of convolutional neural networks and the universal approximation theorem.
2. Apply the theory of training a convolutional neural network.
3. Write programs to design, train and test neural networks to solve problems in computer vision.
4. Competently apply programming and software tools to write programs to solve realistic problems of computer vision using deep neural networks.
5. Understand the principals of testing and evaluating the performance of deep neural networks, and apply these principals in practice.
6. Research an area of computer vision and apply deep neural network methods to a problem in that area.

Research-Led Teaching

This class aims to train students to a stage where they can develop state-of-the-art deep learning for computer vision. In the final project, students will develop an implementation from an existing paper, exploring relevant research and results in their report and presentation.

We will be examining recent literature and recent developments in the advanced topics lectures late in semester.

Examination Material or equipment

Dive into Deep Learning — Dive into Deep Learning 0.16.6 documentation (d2l.ai), Zhang, Lipton, Li, Smola

Deep Learning Book by Goodfellow et al.

Python/NumPy Tutorial by Justin Johnson

PyTorch Tutorial

The Incredible PyTorch - PyTorch Tutorials, Projects, Libraries, Papers etc.

Multiple View Geometry in Computer vision, Book by Richard Hartley

Computer Vision: Algorithms and Applications, Book by Richard Szeliski

Pattern Recognition and Machine Learning, Book by Christopher Bishop

Mathematics for Machine Learning, Deisenroth, Faisal, Soon Ong, available at: https://mml-book.github.io/book/mml-book.pdf

Staff Feedback

Students will be given feedback in the following forms in this course:

• written comments
• verbal comments (including, particularly ask your demonstrator in the Labs)
• feedback to whole class, groups, individuals, focus group etc

Student Feedback

ANU is committed to the demonstration of educational excellence and regularly seeks feedback from students. Students are encouraged to offer feedback directly to their Course Convener or through their College and Course representatives (if applicable). Feedback can also be provided to Course Conveners and teachers via the Student Experience of Learning & Teaching (SELT) feedback program. SELT surveys are confidential and also provide the Colleges and ANU Executive with opportunities to recognise excellent teaching, and opportunities for improvement.

Other Information

The use of Generative AI Tools (e.g., ChatGPT) is permitted in this course, given that proper citation and prompts are provided, along with a description of how the tool contributed to the assignment. Guidelines regarding appropriate citation and use can be found on the ANU library website https://libguides.anu.edu.au/generative-ai

Marks will reflect the contribution of the student rather than the contribution of the tools. Further guidance on appropriate use should be directed to the convener for this course.

Class Schedule

Week/Session	Summary of Activities	Assessment
1	Introduction to deep learning for computer vision	Lab 1
2	Linear classifiers and multi-layer perceptrons	Lab 2
3	Back-propagation and automatic differentiation	Lab 3
4	Image classification and convolutional neural networks	Lab 4
5	Object detection	Lab 5
6	Image segmentation	Lab 6
7	Advanced Topics: debugging algorithms and tools	Project proposal
8	Advanced Topics: recurrent neural networks
9	Advanced Topics: attention and transformers
10	Advanced Topics: selected from contrastive learning, unsupervised learning, vision and language, generative models, deep declarative networks, videos and 3D
11	Advanced Topics: selected from contrastive learning, unsupervised learning, vision and language, generative models, deep declarative networks, videos and 3D
12	Advanced Topics: selected from contrastive learning, unsupervised learning, vision and language, generative models, deep declarative networks, videos and 3D	Project video and report

Tutorial Registration

See Wattle Page.

Assessment Summary

Assessment task	Value	Due Date	Learning Outcomes
Lab 1	5 %	26/07/2024	3,4,5
Lab 2	5 %	02/08/2024	1,2,3,4
Lab 3	5 %	09/08/2024	1, 2, 3, 4, 5
Lab 4	5 %	16/08/2024	1, 2, 3, 4, 5
Lab 5	5 %	23/08/2024	1,2, 3, 4, 5
Lab 6	5 %	30/08/2024	1, 2, 3, 4, 5
Course Project - Group Assessment	70 %	*	4,5,6

* If the Due Date and Return of Assessment date are blank, see the Assessment Tab for specific Assessment Task details

Policies

ANU has educational policies, procedures and guidelines , which are designed to ensure that staff and students are aware of the University’s academic standards, and implement them. Students are expected to have read the Academic Integrity Rule before the commencement of their course. Other key policies and guidelines include:

• Academic Integrity Policy and Procedure
• Student Assessment (Coursework) Policy and Procedure
• Extenuating Circumstances Application
• Student Surveys and Evaluations
• Deferred Examinations
• Student Complaint Resolution Policy and Procedure
• Code of practice for teaching and learning

Assessment Requirements

The ANU is using Turnitin to enhance student citation and referencing techniques, and to assess assignment submissions as a component of the University's approach to managing Academic Integrity. For additional information regarding Turnitin please visit the Academic Skills website. In rare cases where online submission using Turnitin software is not technically possible; or where not using Turnitin software has been justified by the Course Convener and approved by the Associate Dean (Education) on the basis of the teaching model being employed; students shall submit assessment online via ‘Wattle’ outside of Turnitin, or failing that in hard copy, or through a combination of submission methods as approved by the Associate Dean (Education). The submission method is detailed below.

Moderation of Assessment

Marks that are allocated during Semester are to be considered provisional until formalised by the College examiners meeting at the end of each Semester. If appropriate, some moderation of marks might be applied prior to final results being released.

Assessment Task 1

Lab 1

Familiarity with software tools: Python, OpenCV and PyTorch.

Assessment Task 2

Lab 2

Program simple multi-layer perceptron.

Assessment Task 3

Lab 3

Implement a differentiable function in PyTorch and program optimization loop using back-propagation.

Assessment Task 4

Lab 4

Train a CNN for image classification.

Assessment Task 5

Lab 5

Programming task on object detection and evaluation.

Assessment Task 6

Lab 6

Programming task on image segmentation.

Assessment Task 7

Course Project - Group Assessment

Major Research Project. A project proposal, report and presentation are required. This is a group project. Topics will be made available earlier in semester, or you can choose your own topic, subject to discussion with course staff.

Academic Integrity

Academic integrity is a core part of the ANU culture as a community of scholars. The University’s students are an integral part of that community. The academic integrity principle commits all students to engage in academic work in ways that are consistent with, and actively support, academic integrity, and to uphold this commitment by behaving honestly, responsibly and ethically, and with respect and fairness, in scholarly practice.

The University expects all staff and students to be familiar with the academic integrity principle, the Academic Integrity Rule 2021, the Policy: Student Academic Integrity and Procedure: Student Academic Integrity, and to uphold high standards of academic integrity to ensure the quality and value of our qualifications.

The Academic Integrity Rule 2021 is a legal document that the University uses to promote academic integrity, and manage breaches of the academic integrity principle. The Policy and Procedure support the Rule by outlining overarching principles, responsibilities and processes. The Academic Integrity Rule 2021 commences on 1 December 2021 and applies to courses commencing on or after that date, as well as to research conduct occurring on or after that date. Prior to this, the Academic Misconduct Rule 2015 applies.

 

The University commits to assisting all students to understand how to engage in academic work in ways that are consistent with, and actively support academic integrity. All coursework students must complete the online Academic Integrity Module (Epigeum), and Higher Degree Research (HDR) students are required to complete research integrity training. The Academic Integrity website provides information about services available to assist students with their assignments, examinations and other learning activities, as well as understanding and upholding academic integrity.

Online Submission

You will be required to electronically sign a declaration as part of the submission of your assignment. Please keep a copy of the assignment for your records. Unless an exemption has been approved by the Associate Dean (Education) submission must be through Turnitin.

Hardcopy Submission

For some forms of assessment (hand written assignments, art works, laboratory notes, etc.) hard copy submission is appropriate when approved by the Associate Dean (Education). Hard copy submissions must utilise the Assignment Cover Sheet. Please keep a copy of tasks completed for your records.

Late Submission

Individual assessment tasks may or may not allow for late submission. Policy regarding late submission is detailed below:

• Lab assignments cannot be submitted late.
• Late submission of project permitted. Late submission of assessment tasks without an extension are penalised at the rate of 5% of the possible marks available per working day or part thereof. Late submission of assessment tasks is not accepted after 10 working days after the due date, or on or after the date specified in the course outline for the return of the assessment item. Late submission is not accepted for take-home examinations.

Referencing Requirements

The Academic Skills website has information to assist you with your writing and assessments. The website includes information about Academic Integrity including referencing requirements for different disciplines. There is also information on Plagiarism and different ways to use source material. Any use of artificial intelligence must be properly referenced. Failure to properly cite use of Generative AI will be considered a breach of academic integrity.

Extensions and Penalties

Extensions and late submission of assessment pieces are covered by the Student Assessment (Coursework) Policy and Procedure. Extensions may be granted for assessment pieces that are not examinations or take-home examinations. If you need an extension, you must request an extension in writing on or before the due date. If you have documented and appropriate medical evidence that demonstrates you were not able to request an extension on or before the due date, you may be able to request it after the due date.

Privacy Notice

The ANU has made a number of third party, online, databases available for students to use. Use of each online database is conditional on student end users first agreeing to the database licensor’s terms of service and/or privacy policy. Students should read these carefully. In some cases student end users will be required to register an account with the database licensor and submit personal information, including their: first name; last name; ANU email address; and other information.
In cases where student end users are asked to submit ‘content’ to a database, such as an assignment or short answers, the database licensor may only use the student’s ‘content’ in accordance with the terms of service – including any (copyright) licence the student grants to the database licensor. Any personal information or content a student submits may be stored by the licensor, potentially offshore, and will be used to process the database service in accordance with the licensors terms of service and/or privacy policy.
If any student chooses not to agree to the database licensor’s terms of service or privacy policy, the student will not be able to access and use the database. In these circumstances students should contact their lecturer to enquire about alternative arrangements that are available.

Distribution of grades policy

Academic Quality Assurance Committee monitors the performance of students, including attrition, further study and employment rates and grade distribution, and College reports on quality assurance processes for assessment activities, including alignment with national and international disciplinary and interdisciplinary standards, as well as qualification type learning outcomes.

Since first semester 1994, ANU uses a grading scale for all courses. This grading scale is used by all academic areas of the University.

Support for students

The University offers students support through several different services. You may contact the services listed below directly or seek advice from your Course Convener, Student Administrators, or your College and Course representatives (if applicable).

• ANU Health, safety & wellbeing for medical services, counselling, mental health and spiritual support
• ANU Accessibility for students with a disability or ongoing or chronic illness
• ANU Dean of Students for confidential, impartial advice and help to resolve problems between students and the academic or administrative areas of the University
• ANU Academic Skills supports you make your own decisions about how you learn and manage your workload.
• ANU Counselling promotes, supports and enhances mental health and wellbeing within the University student community.
• ANUSA supports and represents all ANU students