Mathematical Image Analysis Group, Saarland University

Information for Prospective Students

Information for Prospective Students Planning to Write Their Thesis with Us

Thank your for your interest in writing your thesis in our group.
Please take a few minutes to read the information below carefully.
Remember to come back to it in the different phases of your thesis.

Prerequisites.
For a bachelor thesis, a high school teacher student thesis, or a research immersion lab, we expect that you have passed our class on Image Processing and Computer Vision. For a master thesis, we additionally assume that you have passed Differential Equations in Image Processing and Computer Vision. The same requirements also hold for requests to co-supervise an external thesis.
Please refrain from sending us e-mails, if we mean this seriously, because we do. These lectures are designed to provide you with the necessary background and familiarise you with our scientific language. Other lectures by us or our colleagues have a different focus and are thus not adequate replacements.
Your specific grades do not matter, but they must be passing grades. However, if we know that you are good, we can offer also somewhat more challenging topics that are closer to interesting open research questions. Also if you have attended additional classes from our group, this may help to offer you a broader scope of topics.
Finding a Thesis Topic.
The most important thing is to find a topic that fits your specific interests in the best way. Therefore, we will be happy to offer you a selection of potential topics that are individually tailored towards your interests. We can offer theses in the areas of visual computing, mathematics, computer science, DSAI, and other programs that allow us to do so. These can be bachelor theses, master theses, theses for prospective high school teachers, and research immersion labs.
If you are interested in writing your thesis in our group, please send an e-mail to Joachim Weickert. You should tell him your background (study program, perhaps also lectures of others that might be relevant for your thesis), which of our classes you have already attended and which are your broader areas of interest.
He will then make an appointment with you and discuss several potential topics with you. Everything is open, and you can choose freely without any obligations. Some topics can be offered by group members, others by Joachim Weickert. The person who offers the topic will be your advisor and your primary contact person.
There is no deadline for the decision about your favourite topic. However, it can happen that some of the topics that have been offered to you are also offered to other students with similar interests. Thus, if you wait too long, you risk that your favourite topic has been taken by someone alse. As soon as we know your decision for an available topic, we reserve it for you, even if you cannot start immediately.

Bachelor / Master Seminar Talk.
Most examination regulations require a so-called bachelor or master seminar talk. In our case this is a short talk (15 minutes) in our group seminar where you present the problem and some first ideas how you would like to solve it. This little extra work is rewarded with quite some credit points.
Please note that the bachelor/master seminar talk takes place in the initial phase of your thesis (a few weeks after you have decided on the topic), not towards the end. It is your responsibility to contact your supervisor in order to fix a time slot for this seminar.
Please e-mail the title of your talk one week before it takes place to Vassillen Chizhov, such that he can announce it via our mailing list.
Your audience mainly consists of MIA group members. You can expect that they are aware of the contents of our regular classes on Image Processing and Computer Vision and Differential Equations on Image Processing and Computer Vision. Please obey the usual rules of good scientific practice that hold for any of our seminars, give fair credits to findings and pictures that do not go back to you (at the same slide), and avoid any plagiarism.
After your talk, please e-mail a pdf file of your slides to Vassillen Chizhov. He will put it on our group seminar website. They will be password-protected and accessible to the MIA group members.
After you have submitted your thesis, you are also expected to present its results to our group in a thesis defence talk (30 minutes, see below).

Registration.
You should also contact your supervisor to register your thesis. This registration must take place at the beginning of the thesis, typically when you give the bachelor / master seminar. We will not accept late registrations shortly before submitting a thesis. The deadlines are for you, not for us: Having a deadline helps you to work in an efficient way.
To register your thesis, please download the form sheet from the examination office website. Fill out everything (also the prospective thesis title and the registration dates), apart from the signature of Joachim Weickert :-). E-mail him the form. It is recommend to state in the form in that you will not use generative AI (see also the next item below). Then Joachim Weickert will sign it and e-mail it to the examination office with you in the cc.
Please keep in mind that once the thesis is registered, the deadline is running. Most study regulations allow 3 months for a bachelor thesis and 6 months for a master thesis. This time span can only be extended for reasons beyond your control (e.g. a longer period of illness). Thus, there is some temporal flexibility before registration, but not afterwards.

Citations and Scientific Ethics.
In science it is extremely important that you acknowledge all sources and give fair credits to other researchers. Failing to do so means that you a stealing results from others. A reader of your thesis must be able to tell what was state-of-the-art from what you have done. If you have received some code or code fragments, this must be clearly stated in your work. If you use scientific results from others, you have to cite them in this context (and not somewhere else).
Literal quotations should be kept to a minimum and must be put in quotation marks. If you use images and drawings from others, you have to state their origin. The same also holds for results generated by AI tools such as ChatGPT, Bard and the like. Since things that do not go back to you do not count as your own results, we disadvise using them.
Please take correct citation ethics very seriously and do not try any plagiarism. There is a high risk that this will be discovered and you fail.

Source Code.
Your code should be written in well-structured C, using not more than 80 characters per line, with English comments. We actually recommend to write the comments before the code. Readability is more important than squeezing the last percent of efficiency out of your code by doing cryptic things. Here is a sample program that provides a number of useful routines for greyscale and colour images: frame.c.
If your thesis relies on neural networks, you should use PyTorch instead of C.
Please include a CD, DVD or some memory card/stick with your code in the copy of your thesis that will be passed on to your supervisor, but not in the other three copies.

Foreign Code.
If you have received code from us, you may not use it for any other purpose than your thesis. In particular, you are not allowed to pass it on to others or to use it commercially.

Meetings with Your Supervisor.
If you prefer regular meetings with your supervisor, let him or her know. If we do not hear anything from you, we assume that you are doing fine.
We recommend to aim at a good balance between being independent and benefitting from the advice and the feedback of your supervisor. Make sure to consult your supervisor in case of problems that you cannot solve by yourself. Please report also major achievements to us to obtain our feedback. On the other hand, please try to show also some independence: If you ask for advise for every little thing that you could have solved easily by thinking about it for a few minutes, you cannot expect a top grade.
If you do not receive a response to one of your e-mails within a few days, this is never a bad intention. Please send it again. In case you are unhappy with some aspects of your supervision, please let us know immediately to give us the chance to improve things.

Appointments.
Supervising many students is a real pleasure, but also time-consuming. Thus, if you make an appointment with your supervisor, make sure to be there in time, since this person has allocated this time slot specifically for you and might be very busy. If you cannot make it to an appointment, let us know as soon as possible, such that this slot can be used for other activities.

Keep it Simple.
Science aims at finding the simplest concept that explains a finding or accomplishes a specific task (Occam's razor). Therefore, you should always try the simplest ideas first. If they work, then it's perfect. Albert Einstein once stated: ``Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genious - and a lot of courage - to move in the opposite direction.''. Also Antoine de Saint-Exupery wrote: ``Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away.''
Before testing your methods with the most complicated real-world images, you should always check first if they meet your expectations with very simple synthetic images. Simple test images are also highly useful to localise bugs in your code. For instance, with a disk-shaped structure you can easily check the rotation invariance of your filter. You will be surprised to learn that we have achieved some of our deepest scientific insights by analysing algorithms for images of size 2 by 2 pixels.

Make Sure That Your Algorithms Have Converged.
We have seen many cases where iterative algorithms have been stopped too early, and the students have drawn inappropriate conclusions from results that have not converged yet. Please do not repeat this mistake. And no, using 100 iterations is not a good stopping criterion. If you solve linear of nonlinear systems of equations iteratively, it is recommended that you make sure that the norm of your residual has decreased by a specified factor, e.g. 0.000001. It is also inappropriate to initialise an iterative algorithm with the correct solution if you know it (e.g. in an inpainting problem).

Good and Bad Results.
It is very important for us to see how you approach scientific problems. As scientists we know that some ideas may fail. Do not worry about this: Even with bad results you can get a good grade if these results are obtained in a thorough scientific way and you understand why things fail. On the other hand, you can also get a bad grade with good results, if you have no intuition and if all ideas go back to your supervisor.

References.
The references in your thesis should be complete and follow a consistent style. Please take this seriously and be thorough. If you use LaTeX, Joachim Weickert we be happy to e-mail you his BibTeX file which contains many references and follows the recommended style.

Time Schedule.
You should reserve plenty of time to write down your results. For a bachelor thesis, four weeks is a typical time span, and for a master thesis we recommend to reserve six weeks. In can be helpful to write down your intermediate results already when you obtain them: Writing down things often helps to discover remaining problems. Towards the end of your thesis, it is better to consolidate and write down your existing results than trying new things that you cannot finish properly.

Draft Versions.
If you want that your supervisor reads your draft before you submit it, this is fine. However, please clarify this early (we might be busy) and allow sufficient time (e.g. two weeks). Moreover, please make sure that these drafts are mature, and do not misuse your supervisor for correcting your typos unless you want to give a bad impression. Please use a large font size (12 point, see below).

Language.
For us it does not matter if you choose English or German, but it should be a language that you master. If you believe that this is German, make sure that you use the correct translation of scientific expressions from English. For instance, "anisotropisch" is not the German word for "anisotropic". In case you would like to pursue a scientific career, it is recommendable to use English, since more people can read your results and it is easier to publish them.

Writing Style.
Regardless of the language, you should write in a clear and direct way and avoid long sentences. Please prefer active language over passive one (e.g. "we choose" instead of "is chosen"), since this sounds less boring. Nobody expects you to write like Shakespeare or Goethe. In particular, there is no need to introduce more than one idea per sentence. Remember Paul Dirac's words: "In science one tries to tell people, in such a way as to be understood by everyone, something that no one ever knew before. But in poetry, it's the exact opposite." Your thesis is about science, not poetry.

Size Does Not Matter.
No, there are no guidelines how many pages a thesis must have. It's the contents that counts. You can get a fantastic grade with a 10-pages work if it is ingenious, and you can fail with a 100-pages thesis if it is of real bad quality.

Font Size Does Matter.
Depending on your study programme, the examination office may have some instructions on the structure of the title page. Indepedently of you study programme, please use a large font size (12 point) in your thesis: In general, your reviewers are older than you, and difficulties reading small font sizes increase with age.

Our Grading Criteria.
Our grade will take into account several aspects, in particular your approach to address a problem in a scientific way, your independence and creativity compared to our input, and your thoroughness and ability to structure and write down your results in a well-readable, logical, and transparent way. Also the coherence of your references (see above) and your thesis defence talk (see below) can influence your grade.
Finding a Second Reviewer.
Most study regulations require that your thesis must be reviewed by two persons (usually professors, postdocs may be possible by asking the head of the examination committee for permission). Please discuss potential candidates you have in mind with your supervisor. Moreover, do not forget to ask this potential reviewer before suggesting his/her name to the examination office.

Extra Copies and pdf File.
Often your direct advisor in the daily business is a Ph.D. student or a postdoc whose research is closely related. In general this person will not get a copy of your thesis from the examination office. Thus, it would be fair and helpful if you give an extra copy of your thesis to him or her. Your direct advisor will use it as a memory of the nice time with you, and as a basis for making suggestions for grading.
Also your parents or grandparents might appreciate an extra copy of the thesis, in particular if they have helped to finance your studies. They will be proud of you. However, don't be shocked if they praise the nice cover more than the contents of the thesis :-).
In modern times with large monitors of high resolution, your reviewers often prefer to read your thesis on their computer. This allows them e.g. to zoom into images. Thus, please send them also a pdf file of your thesis.

Little Pleasures.
Please do not give us presents before or after submitting your thesis, even if you love us and this is common in your home country. Your scientific results are the only way to influence your grade. If you like the way how we have supervised you, simply tell it to your fellow students. That's how you can please us most.

Thesis Defence Talk.
After you have submitted your thesis, contact your supervisor for a time slot for a talk where you present your results to the MIA Group. We are all curious what you have to tell us, and this may also speed up our grading.
The rules for the defence talk are the same as for the bachelor/master seminar, but this time you have 30 minutes instead of 15, and of course you are supposed to present your final results.
In contrast to the bachelor/master seminar, you do not get an extra grade for the defence talk. It counts as part of your thesis work and can influence its grade in a positive or negative way.

Passing Certificate.
In most cases you can get a confirmation from us once you have submitted your thesis and presented your defence talk. It states that we will not grade your thesis worse than 4.0. If you need it before the final grading is completed, just let us know.

Thesis Grading and Review.
Often our stack of work to be reviewed is large. In cases when you need your final grade very urgently, tell us, and we will see what can be done.
If you want, you can of course get the review of your thesis from us. It's about your work, so you may know all about it.

Going on for a Ph.D. Degree ?
A Ph.D. degree demonstrates the ability to work on a high scientific level. It involves several years of research, and it is much more challenging than a master degree. We are constantly looking for strong Ph.D. candidates who satisfy three criteria:
1. The most important point is that you should be creative, work independently, and be able to write down your thoughts in a well-structured, logical and grammatically correct way. An excellent performance in your master thesis is the best opportunity to prove this to us.
2. You should have good technical skills to solve occuring problems. This is documented by good grades. For instance, this is certainly the case if you belong to the top 20 % in our classes on Image Processing and Computer Vision and Differential Equations in Image Processing and Computer Vision.
3. You should fit in our group. This means you should be an open-minded nice person and a team player who has a similar enthusiasm, perfectionism and sense of humor as we have.
When we have the impression that you fulfill these criteria, we will not miss the chance to discuss this with you. While the number of Ph.D. positions in our group is limited, there is also a possibility that we can help you finding a Ph.D. position in another group if your performance is good. This can be in Saarbrücken, somewhere else in Germany, or abroad.

Finding a Job and Staying in Contact.
If you are interested in an image processing or computer vision job in industry, here is a document that also contains a list of machine vision companies in Germany. If you prefer a job in academia, there is also a list of international research groups. We are of course always happy to help you with hints and recommendations. Vice versa, we are also curious about your career afterwards, so please stay in contact.

If you have remaining questions and suggestions, please do not hesitate to contact us.

We wish you much success and look forward to working with you.
The MIA Group