Module COMP52615
Assignment
MISCADA Computer Vision module
academic year 2023/2024
Academic in charge and contact information
Professor Paolo Remagnino, Department of Computer Science
You can reach me by email paolo.remagnino@durham.ac.uk
If a meeting in person is required, send me an email and I will arrange for a suitable date and
time.
Introduction
In this assignment, you are asked to implement software for the detection of a badminton
court and then players on the court. You are then asked to track the players on the court. Data
are provided, read the instructions on how to get hold of the dataset in the dedicated section.
Image frames of a video clip of two international badminton matches have been zipped.
Task specification
A badminton court is defined by a rectangular area delimited by lines. The court in the
available dataset is green, although a court could be of different colour. A badminton court
has a rectangular shape of some given dimensions (see figure 1 for more details), divided into
rectangular regions, related to the game itself.
figure 1: A badminton court, left for single, right for double matches.
You might be a badminton player, perhaps a professional, however, if you are not familiar with
the game, you can find the international rules at https://olympics.com/en/news/badminton-
guide-how-to-play-rules-olympic-history.
We will work with both single and double matches. You are asked to implement solutions to
solve the following tasks:
• Task 1: Image frames detection - for this part of the assignment, you are asked to utilise
both colour segmentation and pattern segmentation to extract all the frames in the
provided video clips that contain the full badminton court. A court is fully visible when an
Module COMP52615
action is shot from the distance, to provide full visibility of a game; this could be from the
side of a court, or from behind one of the two teams. See figure 2 for examples.
Expected Output: Image frames with a fully visible court should be saved in a list of image
names, easily accessible to retrieve the frames for further processing; please do adhere to
the image frame names of the provided datasets.
figure 2: left, single match, from perspective of the player in black
right, a double match, from the perspective of the yellow team.
• Task 2: Court detection - for all the extracted image frames that contain the full view of
the court, you are asked to annotate the court with lines. For this, you will have to develop
code that extracts the lines and effectively builds a 2D model of the court.
Expected Output: both coordinates of the polygons used to highlight the court image
frames with overlapped court should be provided.
• Task 3: Player detection - for this part, you are asked to exploit one of the taught
algorithms to extract bounding boxes that identify the players on the court. If a person is
detected outside the court, its bounding box must be discarded as an outlier.
Expected Output: both coordinates of the rectangles bounding the players and the image
frames with overlapped bounding rectangles should be provided.
• Task 4: Player tracking - for this part, you are asked to use one of the taught techniques to
track the players. To be borne in mind, while the TV camera does not move much from the
best perspective of the game, it might still pan a little, this movement will have to be
compensated for by using the information about the court.
Expected Output: For this task you should provide short video clips using the image frames
and overlapped court lines and bounding rectangles for the players.
Dataset
Data can be found at
https://drive.google.com/drive/folders/15ihQI3D9tU8VsZJsU5sP0FV9JPCpMwr5?usp=share_link
They are two zip files containing image frames of short clips of two badminton matches, a
single and a double, both played at international level. For simplicity, image frames from clips
have been extracted and compressed in zip files.
Module COMP52615
Marking scheme
Task comment %
Description A general introduction to your solution must be provided (5%), as well
as a detailed description of how you solve each of the following tasks
(5%).
10
Task 1 Hint: this seems simple, but it can be tricky, as you have learnt from
the lectures and workshops what appears to be of a given colour, blue
say, is not really blue, so a model of colour in a suitable colour space
must be devised. You might also want to exploit the idea that a court is
well delineated by white lines. Whenever you can, do use combination
of information, with a multitude of features.
Marking: the implemented method must extract all the frames that
have the entire court fully visible, marks will be detracted if some
frames are missed, or some are included erroneously. A 10% will be
assigned to a fully working method, 5% only if the method partially
works.
10
Task 2 Hint: Once the frames of interest have been selected, you then can
search for lines of interest, or shape of interest and find the court
delimiters, the lines defining each side of the court and the various
areas. For this part you can exploit deep learning methods that have
been previously trained to recognise a court.
Marking: the implementation must extract all the lines correctly, so
errors in the detection will result in the detraction of marks. Full 30%
to a complete solution, if the solution works in part, then 15%.
30
Task 3 Hint: For this part I am expecting you to exploit the latest YOLO and
SAM versions to detect the players. The players will have to be in and
on the court, so one or more criteria will have to be defined to discard
any proposals (deep network suggestions of bounding rectangles) that
does not fit. You can also use the idea that at most two players are on
each side of the court.
Marking: the implementation exploits existing deep architectures, so
validation and testing will be essential to get full marks. Ideally, the
result will have to show “tight” bounding rectangles for the extracted
players, with an error measure such as the intersection over union
(IoU) as a viable metric. Comparison of existing architectures applied
to the problem attracts 10%, the other 10% is for the most suitable
metric to assess the result.
20
Task 4 Hint: For this task, you can use the dynamic filters you have been
taught to track players on the court. You can feel free to use any deep
learning method as well you have read during the four weeks.
Marking: the accuracy of the tracking will be a determining factor to
obtain full marks, you are recommended to use the metrics you used
to solve Task 3. Full marks if any spatial-temporal filter is implemented
and manages to track players on the court. A comparison of at least
30
Module COMP52615
two filters attracts 15%. The other 5% is for the generation of a video
clip as qualitative output and 10% for quantitative results.
Total 100
Submission instructions
You are asked to create a single Jupyter Notebook, where you will provide the textual
description of your solutions and the implemented code. Your notebook should be structured
in sections. An introduction should describe in detail the libraries you used, where to find
them and how you solved the tasks. Then you should include one section for each task
solution, where again you describe your solution. The code should be executed, so that the
solutions are visualised, in terms of graphics, images and results; it is strongly recommended
you also include snippets of the formulae implemented by the used algorithms and the
graphics of the employed architecture. Your code should be implemented in Python, using
OpenCV as the main set of computer vision libraries. PyTorch or TensorFlow can be used to
use the deep learning methods you have chosen. Please do make sure your code runs. At the
beginning of your Notebook, provide a detailed description of how you solved the tasks,
including a description of all the libraries you have used and where to find them.
Submission and feedback deadlines
You must submit your solutions by 2pm UK time of the 15th March 2024
Feedback will be provided by 8th April 2024
Extension policy and related documents
If you require an extension for a piece(s) of summative coursework/assignment
you MUST complete the form via the Extension Requests App.
On this form you must specify the reason for the request and the extra time required. Please
also ensure that supporting documentation is provided to support your extension request.
Your extension request will not be considered until supporting documentation has been
received.
If the extension is granted the new deadline is at the discretion of the Programme Director.
Once the completed form has been received by the Learning and Teaching
Coordinator it will be distributed to your Programme Director who will decide on
the outcome. You will normally be notified by email within two working days of the outcome
of your request.
Please note that, until you have received confirmation that your extension request has been
approved, you MUST assume that the original deadline still stands.
Further information on the University's extension policy and academic support can be
found here.
Where circumstances are extreme and an extension request is not enough a Serious Adverse
Circumstances form (SAC) should be considered.
PLAGIARISM and COLLUSION
Module COMP52615
Students suspected of plagiarism, either of published work or work from unpublished sources,
including the work of other students, or of collusion will be dealt with according to Computer
Science and University guidelines. Please see:
https://durhamuniversity.sharepoint.com/teams/LTH/SitePages/6.2.4.aspx
https://durhamuniversity.sharepoint.com/teams/LTH/SitePages/6.2.4.1.aspx
PLAGIARISM with chatGPT or/and similar AI tool
Examples of plagiarism include (but are not limited to) presentation of another person's
thoughts or writings as one's own, including:
- cutting and pasting text, code or pictures from generative AI services (e.g. ChatGPT).
What is acceptable?
If you want to include any text or code produced using generative AI in a submission, that text
or code must be clearly marked as being AI generated, and the generative AI that has
produced the text or code must be clearly cited as the source.
The University has provided a guide on generative AI, with the following page on whether it
can be used in assessment:
https://dcad.awh.durham.ac.uk/generativeai/#/lessons/TW6RxeMtGlc_bRcA2EGVaa_9MrCJ
upK3