INSTITUTE FOR SUSTAINABLE HERITAGE
MSc Data Science for Cultural Heritage
Module BENV0113: Heritage Imaging (15 credits)
Coursework: HERITAGE IMAGING
Overview
You are asked to demonstrate your expertise in some aspect of imaging or image analysis
applied to heritage. You can choose how you would like to demonstrate this expertise.
Examples include:
(1) You can acquire images (e.g, photography, photogrammetry) and be assessed on
the quality of the image capture and subsequent image processing. You will need to
demonstrate expertise in some form of quantitative image analysis (e.g. colour
correction, calibration of scale, image registration, image segmentation, etc).
(2) You can obtain images from elsewhere (even the internet) and devise a procedure
for calibrating and analysing them. For example, you might look at the variation in colour
for multiple images of a particular objects, or you might register images of the same
object taken from different angles, or taken at different times to look for degradation.
(3) We can provide you with a dataset, for example from multispectral or hyperspectral
imaging.
You will be expected to demonstrate the ability to assess image quality, perform realistic image
analysis, describe relevant heritage applications and relate your work to published literature.
Some tips:
 Discuss your choice of image and image acquisition with the module organiser
 Choose appropriate image analysis software (e.g. ImageJ, Matlab, Python) and
demonstrate some expertise
 Show that you understand the relationship between the image and the image quality,
and the image processing that can be appropriately performed upon it.
 Demonstrate your understanding of relevant literature by citing appropriate works by
other researchers.
Please note that this coursework contributes 100% of your grade for this module and you
are nominally expected to spend about 80 hours on it.
老师强调的点：
1. 专业用词要准确不要说模糊的概念“betterquality”之类的要具体说哪方面
2. Researchquestion同理，问题要清晰准确具体，千万不要模糊的问题
3. 要注意对应评分标准的要求，以及guideline，有明确的每个session需要写什么
4. 要把自己的方法和这个field里过去学者的literature作对比分析，弥补了什么gap之类的
5. 根据你的researchquestion去选择用到的软件和编程语言，以及作对应的数据分析和写作
6. 看lecturenotes
7. 最好把时间花在分析解决一个具体的问题上，不要用多个数据集
8. 做到能validate自己的result和方法
以下三种方法都可以，每个都涉及到的lecturenotes里对应的课
1层 λ,
∞
曲
益春
2
Word length and document structure
We strongly suggest that you use the Latex template that you have used earlier. Details here.
The length will be calculated automatically by the Latex template and printed on the front page.
This should not exceed 2500 words. This word limit includes the abstract, main text,
references, titles and subtitles, footnotes and tables, but not figures. Any computer code or
similar can be included as an 'additional file submission' and will not count towards the word
length. Penalties for over-length coursework are given in the Handbook.
Academic integrity
All coursework will be scanned with Turnitin to detect for evidence of plagiarism. Make sure you
are aware of UCL’s rules on plagiarism and the penalties for academic misconduct.
Submission date
The submission deadline is 3rd May 2023 at 15:00 (UK time). The assessment must be
submitted through Moodle as anonymised pdf files.
For submissions which are received after the deadline without pre-approval (e.g. a submitted
Extenuating Circumstances, etc.), various penalties will be applied. Details of these penalties
are published in the Handbook, available here.
Computation facilities
For your coursework, some of you are likely to need facilities for computation. Previous students
have found Google Colab and Matlab helpful and UCL now offers access to the Jupyter python
notebook. You can acquire your own images or download images from the internet, or we can
provide access to some images for analysis. We can also provide some basic web hosting
facilities if that is helpful, though storage is limited.
UCL also offers site licences for some software such as matlab, java, vPython, MS Visual
Studio, ENVI, photoshop, Adobe Creative Cloud, and others such as ImageJ, python, GNU C
and GIMP are freely available.
overleaf的template会给链接！
必须⽤这个哦！
code不用放正文，单独发文件，
正文里解释用到的code就可以
查重很严！引用要非常规范！
编程语言或图像软件不限制！上面提到的都可
以用或者结合用，哪个对于作业更有效更能写
就选哪个，也可以参考给的lecturenotes。但需
要在文章的Method部分里说明选择某个
algorithm/软件的理由，为什么要这么用
∝
∝
谋第∞
∝
←∞毋
∝
春
春越
3
Guidelines on your coursework
Here are some guidelines that might help you with planning and writing your coursework.
Please see also the marking criteria and assessment grid below. Of course, there is more than
one correct way to write a report so I cannot tell you exactly what to do. These guidelines are
suggestions that will probably be appropriate for most, but not all, projects.
Planning your coursework project
1. Research statement
I strongly suggest that you develop a clear statement of your research question and state this
explicitly in your report. This statement should be clear and answerable in an objective way.
One way to phrase this is as a hypothesis, which is a testable statement.
For example, “photogrammetry with photographs taken from a phone camera gives good quality
models” is a poor hypothesis as you don’t define what is meant by “quality” and there’s no real
way to test the hypothesis. “photogrammetry with photographs taken from a phone camera
gives better models than photogrammetry with photographs taken from a professional camera”
is better as you are starting to define how you do your test, but “better” is undefined.
A good hypothesis might be “photogrammetry with photographs taken from a phone camera
gives a surface mesh that is indistinguishable from meshes obtained using photogrammetry with
photographs taken from a professional camera”. This is testable, objective and all the terms can
be clearly defined.
A research statement in itself might not make a great change to your project, but if you can
clearly define a single, clear, testable research question, you can use that to inform your
selection of images and techniques, as well as your data analysis and writing.
2. Your work
Data science covers all aspects of the imaging pipeline from image acquisition through to
analysis. Explain where your images came from and why you chose them. If you took the
photographs yourself, write about that. How did you choose the camera settings? If you got
them from elsewhere, why do you think they are reliable? It’s only meaningful to do high-quality
data science if you believe the data is good quality.
3. Scope
It is easy to begin by being too ambitious. Manage your time so that the effort you put into this
coursework is appropriate. Don’t do unnecessary work. One carefully planned and well thought
out experiment is likely to lead to a better project than multiple comparisons between different
datasets. It’s probably best to focus on answering one question in detail than trying to answer
many minor questions.
4. Level of understanding
Assume you are writing for the other students in your year group. They have studied the same
modules as you, but don’t have the specific understanding of your project. In the introduction,
you need to explain any additional material that hasn’t been covered in the taught modules that
a reader would need to know in order to understand your project.
5. Validation
You will receive credit for high-level analytic skills, as expected at Masters level. One way of
showing these skills is to show you understand how to validate your results. This means being
able to persuade a reader that your results are correct. Maybe you need to compare your
method with control examples (for example, if you are examining the effect of compression on
image quality, you might compare images before and after compression). Designing and
carrying out robust, objective validation is a good way to get the credit needed for a distinction.
一定要看评分标准！会有比较
具体的要求
一个好的researchquestion例子，不可以用这个
但是要一个类似这样清晰具体的问题
用到lecturenotes里没有提
过的方法时要解释一下
4
5. Discussion
This is where you can demonstrate your deeper understanding of data science. We want to see
good quality data analysis, but also an understanding of data quality throughout the project. It is
important to consider why you chose the dataset you did, and why you chose the data analysis
methods. Be reflective.
Writing your report
6. Abstract
As you read more scientific papers, you’ll notice they have an abstract. This is a summary of the
method, results and main conclusions of the paper. Think of it as an advertisement, helping you
to decide whether to read the paper or not. Please include an abstract in your report. It will show
that you have identified the important parts of your work, and will help the assessor to
understand your project.
6. Structure
It is almost always the case that the standard report structure is best: abstract, introduction,
methods, results, discussion, conclusion. In some circumstances, you might choose a different
structure, but you should then think about why the standard structure is not appropriate. In the
introduction, you describe the research question, say why it is important, provide any necessary
background information and give relevant references. Your hypothesis or research statement
might come at the end of the introduction. In the methods, you will write about the data science
algorithms you used and how you chose any training set, test sets or experimental sets of data.
The results will include the data you have created during your project, with relevant images,
figures, graphs, tables and so on. The discussion is where you write in detail about your
interpretation of the results, reflect on the implications of your results and describe how your
work relates to published literature.
7. Writing style
Your writing should be concise, precise and clear. 2500 words is a fairly short report. Writing
concisely is a high-level skill that we will be testing in this project. Being precise means
choosing the correct words to explain a concept. For example, if you are comparing two images,
words like “best”, “quality” and “clear” are rarely precise. Instead, you should normally use terms
like “exposure”, “spatial resolution”, “signal-to-noise” and so on. Finally, clarity means that your
writing is easy to understand. Sometime it is hard to be both concise and clear at the same time
and you need to find a compromise.
8. Definitions
Complex or ambiguous terms should be defined (or, better, avoided!)
9. Computer code
You may include computer code as an appendix. However, it would be better to describe the
code in the methods section. Explain why you chose that particular algorithm. Why did you
implement it in that way? Why did you choose that language? How did you test your code?

Finally, remember that you are studying for a qualification in Data Science for Cultural Heritage:
you should demonstrate your knowledge of data science in this area.
一个清晰的abstract
要有反思性思维优点，不足，futurework
文章包含的6个part
intro的内
容
methods
Results
discussion
语言准确清晰简洁，不要有语法错误，用专业词汇
尽量规避模糊用词
×
±
5
Marking Criteria
Mark Numerical
mark
Criteria
A+ 90-100
D
is
tin
ct
io
n
Immediately publishable.
Original contribution to the field, clear, excellent
argument and use of non-textual elements. An
extensive, thorough and wide ranging literature study,
with powerful conclusions.
A 80-89 As for A+, but with a few minor corrections.
A- 70-79 As for A+, but with minor corrections.
B+ 67-69
Pa
ss
Potentially publishable, corrections needed.
There are original elements to the literature review,
the text flow is clear. The style is predominantly
analytical. Non-textual elements might require further
work, but are relevant. The conclusions are relevant.
Minor corrections are still needed.
B 64-66 As for B+, but with a few major corrections.
B- 60-63 As for B+, but with major corrections.
C+ 57-59 Possibly publishable, major corrections needed.
There are few original elements to the review. The
style is predominantly encyclopaedic/descriptive. Non-
text elements require further work, or are redundant.
Many language errors. The argument lacks clarity and
the conclusions could be stronger.
C 54-56 As for C+, but with several substantial corrections.
C- 50-53 As for C+, but with many substantial corrections.
D-F 0-49
Fa
il
Not publishable even after many major
corrections.
The aims are unclear, the non-textual elements are
redundant, not intelligible, or unclear. The text is
encyclopaedic/descriptive, with no original
contribution. The language is unclear, with numerous
errors.
6
Heritage Imaging Assessment Grid
Technical content Excellent Distinction Merit Pass Near Pass Fail
Evidence of understanding:
Understanding of imaging practice: image as part of a pipeline,
image quality, image processing, image analysis etc
Convincing expertise in the
subject area.
Demonstrates good
knowledge of the main
topics.
Demonstrates good
understanding and some
original insights.
Demonstrates fair understanding.
Some understanding is
demonstrated.
Superficial understanding,
major gaps in basic
concepts.
Critical analysis:
Demonstrates insights on how results might be improved.
Appreciation of strengths and limitations of methods used, and
their impact on results. Insights into use of imaging in heritage
Excellent demonstration of
insight. Persuasive
conclusions.
Good presentation of
results and good insight.
Strong conclusions.
Some relevant synthesis of
results and some insight. Some
sound conclusions.
Some discussion of results, with
incorrect interpretation. Weak
discussion of methodology. Some
valid conclusions.
Attempt to synthesise
results, and a basic
discussion. Some valid
conclusions.
Poor or insufficient attempt
to synthesise results.
Inappropriate or no
conclusions.
Experimental
 Design and implementation of the imaging method.
 Experimental data collection and analysis.
Demonstration of non-
standard imaging methods
Imaging methods go
beyond standard
methods.
Excellent demonstration of
standard imaging methods.
Acceptable practical imaging
technique with clear explanation
of methods
Some attempts made, but
imaging poor
Negligible contribution.
Computational
 High quality, well focused image processing and analysis
Clear explanation of novel
image processing.
Potentially publishable.
Some non-standard
methods used, or
standard methods used
in novel ways. Well
explained.
Standard processing and
analysis performed accurately
and explained clearly.
Standard image analysis, perhaps
with errors.
Weak attempt at carrying
out and explaining image
analysis
Unacceptably poor attempt
at image analysis.
Structure, writing & presentation
Structure:
 Organization; use of paragraphs and sections to present
ideas. Use of sections to make text easier to read
Excellent. Good. Coherent structure, but some
ideas/arguments are not
concise or are redundant.
Many ideas/arguments are not
concise or are redundant and
disrupt the flow.
No coherent structure.
Ideas/arguments are
superficial and redundant.
No apparent structure.
 Statement of aims and outcomes of the project. Clear. Clear. Are stated within the report. Evidence of reflection but not
clearly presented.
Weak evidence of
understanding.
No evidence of
understanding.
Microstructure:
 Writing style is clear, succinct, easy to follow, and engaging.
Extremely good style
throughout report.
Mostly extremely good
style.
Generally good style for most
paragraphs.
Some paragraphs superfluous or
difficult to follow.
Poor style. Occasionally
paragraphs lack structure.
Poor style throughout
report.
 The information provided is relevant, concise, and non-
repetitive.
Excellent throughout. Excellent for most text. Generally good for most
paragraphs.
Average. Little or no summarized
information.
Does not convey main
ideas/arguments.
Unstructured information
 Language is formal and scientific. Complicated sentences are
avoided. Technical terms defined clearly and used
accurately
Correctly used. Correctly used in most
text.
Mostly formal but some
occasional jargon.
Alternates between formal and
colloquial.
Colloquial language. Colloquial language.
Presentation:
 Presentation quality. Use of figures, graphs etc.
Publishable quality. Near publishable quality. Satisfactory quality. Average quality. Poor quality. No consideration of
presentation.
Literature review & referencing
 Evidence of a wide-ranging research-based literature review,
and appropriate reference to the leading and most relevant
work in the area.
Excellent wide-ranging and
relevant literature review.
Excellent wide-ranging
and relevant literature
review.
Knowledge, analysis and
evaluation of a range of mainly
relevant literature.
Some relevant literature, but
limited and not always relevant.
Little or no logical relevant
literature apart from
recommended sources.
Little or no appropriate
references, even to
recommended sources.
 Evaluation of the literature and exploration of the research
question(s) in a balanced and comprehensive manner.
Excellent, critical insight.
All major ideas presented.
Good insight. Most
major ideas presented.
Fair evaluation for parts of the
research question.
A fair synthesis but poor critical
discussion.
Inaccuracies, and lacks
critical analysis/discussion
of literature.
Many inaccuracies and no
critical discussion of
literature.