Evaluate and apply a range of machine learning algorithms and the relevant theories, concepts, principles, and practises commonly used within the image processing and computer vision problem domain.

2024-12-02 15:21:42

Assessment Brief

Module Leader:		Level: 7
Module Name: Machine Learning and Computer Vision		Module Code: 55-708255
Assignment Title: Computer Vision System Development
Type: Individual Work	Weighting: 100%	Magnitude: 10 mins demonstration
Submission date/time: O5 Decmber 2024	Blackboard submission: Yes Turnitin submission: No	Format: source code
Planned feedback date: 3 Working Weeks	Mode of feedback: Summative Feedback will be given after demonstration in written Formative feedback will be given during support session
In this assessment are students asked to consider:	Inclusivity and accessability		Not applicable
	Sustainability		Not applicable

Module Learning Outcomes Evaluate and apply a range of machine learning algorithms and the relevant theories, concepts, principles, and practises commonly used within the image processing and computer vision problem domain. Evaluate and apply the methods and techniques required for processing image and video signals. Identify and apply relevant artificial intelligence techniques to solve practical problems in key industrial fields. Design and implement appropriate solutions for computer vision tasks using various tools, methods and techniques based on data-driven approaches. Critically evaluate those approaches by measuring system feasibility and flexibility when facing challenging real-world scenarios

Assessment Brief

Introduction

This assignment aims to assess your problem solving, system design/development, and critical evaluation skills while creating and testing a computer vision system.

This is an individual project. You need to create a handwriting recognition system which is capable of distinguishing specific handwriting numbers (0-9) from a group of given images and videos captured by scanned handwriting from a given test dataset. You need to use computer vision and machine learning skills for this task.

You are expected to create your own end-to-end machine learning and computer vision algorithms by using data- driven approaches (i.e., image pre-processing, feature analysis, learning algorithms, and performance validation). You should aim at improving your system performance with multiple experiment settings and variable system evaluation approaches.

The weight of this course work is 100% and you need to prepare the following contents for the assessment:

All your source code of this project. You may have multiple versions of your code (especially when you have many different experiment settings). You are expected to submit all the used and unused code for the assessment.
A 10-minute demonstration with a tutor. During the demonstration, you are expected to explain your code`s function, analyse your results and raise discussion points with the tutors. The tutors will also ask questions about the algorithms` functionality and the system`s performance.

You will have opportunities to show all or part of your work before the submission deadline and receive formative feedback from tutors.

After submission, your tutor will send booking invitation emails for your demonstration. The demonstration will be scheduled between Week 11 and Week 12 (i.e., from December 2–13, 2024) of the teaching period. Ensure you secure a slot for yourself, as no marks will be awarded without a demonstration.

Getting Started

You are recommended to use MNIST dataset and some suitable machine learning algorithms to train your handwriting recognition algorithm.

The test image/video samples of scanned handwritten numbers are provided for this assignment. You need to download the data from Blackboard at “Assessment” → “Task: Computer Vision System Development” → “Test Data”. You can also use some of your own handwritten numbers and take a photo of them for testing your algorithms.

You can use any IDEs, such as PyCharm, or Google Colab, to create the software artifact. However, the image processing, computer vision, and machine learning algorithms must be created by yourself without using high- level off-the-shelf libraries (i.e., You need to demonstrate that you know how to implement the underlying algorithms and data structures). Please talk to your tutors if you are uncertain which functions can be used in the implementation.

If you need to use other development toolkits, especially some commercial software, products, or licenses. please discuss the ideas with the tutors.

If there are aspects of the implementations that you are unable to code, then you should create a placeholder for them and comment out the code that is causing issues to ensure the project can run.

Recommendation

Once you create your base-system, you are encouraged to carry out more exploration from there and aim at creating a better object detection application. You can either use taught and untaught techniques for this project. You are welcome to add new functions and improve the source code. You can discuss the ideas if you have specific additions in mind.

Some of the recommendations are:

A. Image Feature analysis

You can use variable image processing and feature extraction algorithms to get the best image feature for the machine learning system. You can also measure and analyse feature and target dependencies for the tasks. In addition, you can try to use some projection or embedding techniques for the visualisation such as Principal Component Analysis (PCA), or t-Distributed Stochastic Neighbour Embedding (t-SNE) for your feature analysis tasks.

B. Machine Learning

Design and develop a suitable learning model using machine learning algorithms and suitable data structures. You may need to pre-process the data such as "scaling" and "de-noising". You can use cross- validation approaches to estimate the performance. You also need to keep the balance between under- and over-fitting through adjusting the hyperparameters of the chosen machine learning algorithms.

C. Evaluation

The focus will be the accuracy performance of the developed system. You need to fine-tune your model to improve system accuracy performance. It would be better if you can compare more than two different

algorithms in the project. During the system evaluation, you need to analyse and discuss the performance by using some scientific evaluation approaches such as ROC curves and confusion matrices.

You should attach a reference list in your code comments for anything not originally created by yourself, such as existing code from GitHub, and Python code from Generative AI tools such as ChatGPT, Copilot, and Bard.

If you encounter any issues with your code not functioning correctly, please make sure to comment out the errors and submit the entire code for assessment.

Submission

Your code should be submitted electronically through the module`s Blackboard site as a single ZIP file that contains all your material before the deadline 5^th December 2024, at 15:00). If for any reason you cannot submit your assignment to the Blackboard, then you should copy all the deliverables to a Memory Stick and hand it into Cantor 9312 by the same deadline.

The coursework sent as email will not be marked.

If you create a large file or dataset (more than the file limit of the Blackboard site), you should add a Google Drive link to the documentation. Make sure the time stamps in the shared drive are earlier than the deadline. Do not submit the original test images to the module website.

Who do I contact if I have a question?

Ask the tutor most closely related to the issue first. If you don’t feel the matter is resolved, then ask the module leader/course leader (Jing Wang, jing.wang@shu.ac.uk). If you still don’t feel it’s resolved, ask the Deputy Head of Computing (Mark Jacobi, m.jacobi@shu.ac.uk).

Please note that marks are decided by the teaching team, are verified internally, and not debatable, but your tutors will be happy to explain the feedback you have been given. Please let your tutor know straight away if you think there’s been a mistake in any assessment procedures

Artificial Intelligence and Academic Integrity – AI&AI

It is important you do not use AI tools to generate an assignment and submit it as if it were your own work. Our regulations state:

Contract cheating/concerns over authorship: This form of misconduct involves another person (or artificial intelligence) creating the assignment which you then submit as your own. Examples of this sort of misconduct include: buying an assignment from an ‘essay mill’/professional writer; submitting an assignment which you have downloaded from a file-sharing site; acquiring an essay from another student or family member and submitting it as your own; attempting to pass off work created by artificial intelligence as your own. These activities show a clear intention to deceive the marker and are treated as misconduct.

Further guidance is available here: https://blogs.shu.ac.uk/assessment4students/preparing-to-submit-work/#AI

Assessment Criteria

University Grade Descriptor (UGD) for Level 7 (Postgraduate Level)

Class	Category	Mark range	%	General Characteristics
Distinction	Exceptional Distinction	93 - 100	96	Exceptional breadth and depth of knowledge and understanding evidenced by own independent insight and critical awareness of relevant literature and concepts at the forefront of the discipline; evidence of extensive and appropriate independent inquiry operating with advanced concepts, methods and techniques to solve problems in unfamiliar contexts; Cogent arguments and explanations are consistently provided using a range of media demonstrating an ability to communicate effectively in a variety of formats using a sophisticated level of the English language in an eloquent and professional manner to both technical and non-technical audiences; a sustained academic approach to all aspects of the tasks is evidenced; academic work extends boundaries of the disciplines and is beyond expectation of the level and may achieve publishable or commercial standard.
Distinction	High Distinction	85 - 92	89	Excellent knowledge and understanding evidenced by some clear independent insight and critical awareness of relevant concepts some of which are at the forefront of the discipline; evidence of appropriate independent inquiry operating with core concepts, methods and techniques to solve complex problems in mostly familiar contexts; Arguments and explanations are provided that is well-supported by the literature and in some cases uses a range of media demonstrating an ability to communicate effectively in a limited number of formats using own style that is suited to both technical and non-technical audiences; a sustained academic approach to most aspects of the tasks is evidenced; one or more aspects of the academic work is beyond the prescribed range and evidences a competent understanding of all of the relevant taught content.
	Mid Distinction	78 - 84	81
	Low Distinction	70 - 77	74
Merit	High Merit	67 - 69	68	Very good knowledge and understanding is evidenced as the student is typically able to independently relate taught facts/concepts together some of which are at the forefront of the discipline; evidence of some competent independent inquiry operating with core concepts, methods and techniques to solve familiar problems; Arguments and explanations are provided that are typically supported by the literature and in some cases may challenge some received wisdoms; competently uses all taught media and communication methods to communicate effectively in a familiar settings; an academically rigorous approach applied to some aspects of the tasks is evidenced; some beyond the prescribed range, may rely on set sources to advance work/direct arguments; demonstrates autonomy in approach to learning.
	Mid Merit	64 - 66	65
	Low Merit	60 - 63	62
Pass	High Pass	57 - 59	58	Satisfactory knowledge and understanding of the area of study balanced towards the descriptive rather than critical or analytical and mostly confined to concepts that are not at the forefront of the discipline; evidence of some independent reading and research to advance work and inform arguments and approaches; Arguments and explanations are limited in range and depth although some are adequately supported by the literature albeit descriptively rather than critically; competently uses at least one taught media and communication method to communicate appropriately in familiar settings; although the approach applied to some aspects of the tasks may lack academic rigour, there are some clear areas of competence within the prescribed range. Relies on set sources to advance work/direct arguments and communicated in a way which shows clarity but structure may not always be coherent.
	Mid Pass	54 - 56	55
	Low Pass	50 - 53	50
Fail	Borderline Fail	40 - 49	45	Knowledge and understanding is insufficient as the student only evidences an understanding of small subset of the taught concepts and techniques; fails to make sufficient links between known concepts and facts to adequately solve relevant aspects of the brief/problem; little ability to independently select and evaluate reading/research with almost total reliance on set sources and unsubstantiated arguments/methods; communication/presentation may be competent in places but fails to demonstrate structure, clarity and/or focus; inability to adequately define problems and make reasoned judgements; the general approach to tasks lacks rigor and competence.
	Mid Fail	30 - 39	35
	Low Fail	20 - 29	25
Fail	Very Low Fail	6-19	10	Knowledge and understanding is highly insufficient as the student is unable to evidence any meaningful understanding of taught concepts or methods; very limited evidence of reading and research to advance work; inadequate technical and practical skills as the student is unable to use and apply such skills to address problems or make judgements; limited or lack of understanding of the boundaries of the discipline and does not question received wisdom; approach to learning lacks autonomy and approach to tasks is not sustained; inability to communicate coherently.
Zero	Zero	0-5	0	Work not submitted, work of no merit, penalty in some misconduct cases.

100% Plagiarism Free & Custom Written, Tailored to your instructions