ECS7027P: MACHINE LEARNING DEPLOYMENT
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
ECS7027P: MACHINE LEARNING DEPLOYMENT
Resit Assignment
(worth 100% of total mark):
Facial Emotion Recognition through
Machine Learning
Description
The assignment is designed to investigate the problem domain of Facial Emotion Recognition (FER) in real-life environments, i .e ., in-the-wild and demonstrate your understanding on data preparation, data pre-processing and developing machine learning approaches for FER .
The dataset needed for this assignment can be downloaded fromhere. All images in the provided dataset are annotated in terms of the 6 basic emotions (i.e., happiness, surprise, disgust, fear, sadness, anger) and a category named ‘other’ that involves affective states other than the 6 basic emotions.
Tasks
1) Conduct a literature review/survey regarding the studied topic. Three recent (i.e., within the last 5 years) research works should be referenced, mentioned, discussed and appropriately and critically analyzed. This includes defining the exact scope of the paper, mentioning the dataset utilized, the proposed methodology and its novelty compared to other works and the presented results and their limitations. (18 Marks)
2) Perform and explain appropriate data preparation techniques on the provided dataset. Results-statistics (before and after these techniques have taken place) should be presented. (15 Marks)
3) Which are two common data pre-processing techniques found in the literature and being applied when the dataset involves people’s faces (in other words which two common pre-processing techniques are applied for FER which are specific to FER and the face analysis domain in general, but they are not applied to any other domain)? Perform all necessary and appropriate data pre-processing techniques to the provided dataset and explain accordingly. (25 Marks)
4) Develop a machine learning model that solves the studied problem. Illustrate the model’s performance on the training and validation sets of the dataset using appropriate performance evaluation metrics. What is the performance of the model on the test set? Explain, discuss and critically analyze the obtained results. If the results are not good-adequate enough, explain why this is happening and suggest and perform strategies to alleviate such issues. (27 Marks)
5) Perform an ablation study; this means that you need to pick 3 hyperparameters of the model and tune them and show results (e.g., figures, tables, plots) and how they influence the model’s performance; explain, discuss and critically analyze the illustrated results. (15 Marks)
Notes:
1) each action taken should be justified
2) this is an individual assignment
Submission Requirements - Deliverables
Write a report about what you have done, along with relevant plots (addressing the tasks mentioned above). The report should be at maximum of 7000 words (excluding references) and should follow the provided latex template(it is recommended to use overleaf – just create an account, then create a blank project and upload the provided latex template) . The latex structure (e.g., margins, font, font size etc) should not be changed. The report should be in .pdf format named as: ECS7027P_Resit_StudentID .pdf (replace ‘StudentID’ with your student ID).
Only a written report should be submitted.
Marking Criteria
• Clearly and succinctly written report
• Sufficient explanations, comments, diagrams that answer the questions and show understanding .
Writing Style
The paper should be written following an academic writing style (formal prose articulating a rigorous reasoning using evidence and references to other academic works, and prioritising such reasoning over emotional feelings). The paper should demonstrate critical thinking. All works from the literature as well as the references of the provided dataset should be included in the list of references.
2023-01-19