Image Classification using Deep Learning Computer Science homework help


CSY3025 Artificial Intelligence Techniques

Assignment 1 – Image Classification using Deep Learning

The Definitive Guide to Reading Facial Microexpressions

Sentiment analysis through the detection of facial expressions is a useful tool to help understanding consumer mood, emotion, or the degree of concentration in real-time. For instance, car manufactures can use this tool to detect whether a driver is stressed, angry or tired. Teachers can use emotion recognition as a feedback of student engagement. Future computer games can also be designed to adapt game content based on player emotions in order to improve the gaming experience.

Your task is to design and implement an image classifier that can recognise emotions (categories of facial expressions) based on close-up facial images captured from cameras. As a minimum requirement (to support a passing grade), the classifier should have the capability of handling at least 5 (five) different emotions with a reasonable performance. You must also critically analyse any bias in your dataset and model design. To achieve a higher grade, you are expected to design and implement additional features. Examples of additional features are (but not limited to): additional emotion categories, inclusion of different ethnicities, exceptional performance through “in-the-wild” testing, implementations of user interfaces / user applications, data visualisation, comparative analysis of alternative solutions, etc.

To complete the task, you must study the problem space, establish image dataset, design and train a deep learning classifier, and use appropriate methods to evaluate the performance of your classifier.

The deliverables of the assignments include:

1. A working deep learning model implemented using Keras/TensorFlow and Python. You must submit a Python notebook file (.ipynb) with all outputs saved.

2. A report that includes but not limited to the following sections:

a. Cover page

i. A link to your dataset (e.g., github)

b. Introduction

c. Problem analysis and background research

d. Building deep learning network (explain how you make choices at each following step)

i. Dataset

ii. Network (structure, loss function, optimiser, etc.)

iii. Training and evaluation

iv. Testing

e. Summary of additional features (if applies)

f. Discussions and conclusions

i. Lessons learned

ii. Limitations

g. References

3. A 5-minute demo video about your design/implementation

There will be three submission points:

1. Report

2. Source code (Python notebook and application if applicable)

3. Demo video

The marking will be based on:

1. 20% Problem analysis and background research (demonstrate your general knowledge of deep learning and the application scenario)

2. 30% Design (Architectural design of your application, construction of dataset, test plans, etc. including additional features such as extra categories, exceptional model performance, user interface, etc.)

3. 40% Implementation (completion of the model, reasonable performance, and additional features)

4. 10% Report quality (quality of technical writing)


Levels of Achievement



Very good


Needs some more work

Needs much more work

No submission

Problem analysis and background research (20%)

14 to 20 points

Excellent analysis of the problem space and requirements. Research of related applications and deep learning frameworks with references to reputable sources. Demonstrate how the research outcomes feed into design.

12 to 13 points

Good discussions of the problem space. Some research of the related topics clearly referenced.

10 to 11 points

Problem analysed and summarised. Shows some understanding of the topic. Some reference included.

8 to 9 points

Limited problem analysis and summary. Minimum background research

1 to 7 points

Very little work on research.

0 to 0 points

No covered

Design (30%)

21 to 30 points

Outstanding architectural design of your application, construction of dataset, test plans. Relevant and useful additional features such as extra classes, exceptional model performance, experimentation with different models, user interfaces, etc. Clear design diagrams.

18 to 20 points

Very good design with some additional features.

15 to 17 points

Basic design that fulfils minimum requirements.

12 to 14 points

Some errors in design

1 to 11 points

Major issues in design

0 to 0 points

No covered

Implementation (40%)

28 to 40 points

Excellent implementation of a deep learning based application. Outstanding model performance. Reliable commented coded. All design realised and fully functional. Many useful additional features. Good user experience and code quality. Usability and/or user experience testing.

24 to 27 points

Working solution with well coded additional features.

20 to 23 points

Working solution with no additional features.

16 to 19 points

Some minor issues in implementation

1 to 15 points

Major issues in implementation

0 to 0 points

No covered

Technical writing (10%)

7 to 10 points

Excellent structured document that makes a good use of figures, tables, reference, etc. The report is easy to follow and includes sufficient details. Critical conclusions including limitations and proposed improvements.

6 to 6 points

Good technical writing and report structure that is easy to follow.

5 to 5 points

Satisfactory writing with most key details.

4 to 4 points

Difficult to read and lacks details

1 to 3 points

Hard to read and lacks details

0 to 0 points

No covered

Looking for this or a Similar Assignment? Click below to Place your Order