Trinity College Dublin

Project organisation.

While the projects that I supervise are all individual they generally all try to solve some kind of (somewhat novel) application problem in the area of computer vision. I put forward a number of proposals each year but strongly encourage people to propose their own projects in areas which interest them (as I have often found this to result in better projects!).

Weekly Meetings.

Stages of project development.

1. Initial project specification. Where projects have been proposed by me, I ask students to write a specification for the project based on their understanding of what is intended on the basis of a conversation we have had about the project. This is to ensure that we are on the same page. Students proposing their own project concepts need to do this as well, but it is based more on expanding what they have initially proposed. All the project specifications need to consider not just what the project concept is, but also look at matters like testing (e.g. how will the system be tested, where will the test imagery come from, will ethical permission be required).
2. Background researchThe projects will invariably require research in the relevant field looking for similar work particularly in the academic literature). When reading papers students need to take brief notes (so they won't have to read them again when writing the background chapter of their thesis) and should give brief summaries to me. Some resources for locating papers:
   • Google Scholar.
   • IEEE Xplore.
   • CiteSeerX.
3. Understanding the goal. It is important to have a feel for the goal (the report/thesis/dissertation) early on in the project and to help with this I make some of my reports from previous projects available to you. You should use these to give you an idea of report structure, quality of diagrams, type of writing (impersonal) and the quality of references.
4. Project specification and project plan. Once the background research is completed the project specification needs to be made more concrete, preferably with a series of goals (which increase in complexity) allowing for success to be demonstrated at different stages. A plan/schedule for the project work needs to be created at this stage, with as much detail as possible. It would be expected that this will need many revisions, but at least provides a context for the work which will hopefully help keep students on track.
5. Data acquisition and Ethical approval. One critical aspect of a project in computer vision is the acquisition of imagery for development and testing. This may come for an online repository or may need to be recorded. In the later case this is probably going to need ethical approval which can take a while to obtain. To assist with the application process I have collected a number of successful ethics applications from previous students which can be used to help craft a new application. Ethics applciations is made online in Trinity using a system called REAMS (click here).
6. Ground Truth and Testing. Almost all students have to create ground truth for the images/videos which they have acquired so that the accuracy of the developed system can be judged during the testing stages. Time will need to be spent developing a test environment so that the system can be automatically tested against a wide range of images/videos easily. This test environment should be developed very early on in the lifecycle of the project so that feedback can be obtained regularly on development work. You should avoid just testing on single images/videos and visually evaluating the results.
7. Development work. This is where the main body of work is done but it only makes sense if done in context with all the other aspects of the project. Some basic computer vision resources:
   • OpenCV.
   • Image Processing Learning Resources (Edinburgh).
   • CV online - An On-Line Compendium of Computer Vision (Edinburgh).
   • Geeks for geeks. Vision, Deep learning, Deep Learning for Vision, ...
   • OpenPose.
8. Progress Report. On some degree programs (such as Engineering), students are required to complete an interim progress report which is graded. Typically these are due around January sometime. Students need to check the requirements and deadlines imposed by their own degree programmes.
9. Demonstration. Around 2-5 weeks before the thesis submission deadline, students are required to make a presentation and demonstration to their supervisor and second reader. This allows the student to get the project concept (and execution) into the mind of the examiners and gives the chance for examiners to request that particular aspects are highlighted in the thesis. For some degree programmes these demonstrations are marked and count towards your final grade.
10. Writing the thesis. For must students the thesis is the longest document they have ever written. The first thing to do is to study some previous theses and many of the theses which I have supervised in the past are available online. Some important aspects to note are the structure, the use of diagrams, and the quality of the references. The background chapter can be written very early on in the lifecycle of the thesis which will give you good practice of writing. Please note that a good thesis is not written by just writing it and submitting it - there should be a number of iterations where unnecessary text is deleted and many sections are re-written to make them clearer or more concise. A lot of time should go into the creation of good diagrams/images/charts. Also bear in mind that the thesis should be directed at those examining it - telling a single story describing the system you have specified, built, and tested. That said it is not a diary of what you did.

Evaluation.

• Problem statement, motivation, and analysis (10%)
• Background Research & Literature Review (15%)
• Technical content and project execution (50%)
• Testing, evaluation, critical analysis & conclusions (15%)
• Report presentation and writing (10%)