ML6 Thesis Projects
As ML6, we focus on cutting-edge technologies. This can only be done by investing time in research. While our ML6 agents are doing research, they sometimes discover interesting topics. Unfortunately, an ML6 agent doesn't always have enough resources to investigate these topics in-depth.
That's why ML6 is opening up parts of their research in the form of theses. As a thesis ML6 agent you'll be guided by an expert on the topic and will be able to enjoy the full ML6 experience, while gaining valuable experience using the latest technologies in a fast-paced environment.
You will be a full member of our team and work on one of the following internal projects as a Machine Learning Engineer or Data Engineer. Keep a close eye on this page as we are continuously adding new exciting projects across our different locations.
Apply now (the 'internship' job opening entails both internships and thesis research projects).
1. Transfer Learning in Time Series
Transfer Learning - Time Series - Machine Learning - Artificial Intelligence
As an AI consultancy firm, we want to build real-world applications of Machine Learning as efficiently as possible. One way to cope with the small amount of data that customers often have is to use transfer learning. Transfer learning is already applied in computer vision and in NLP very often, but in time series applications, there is no go-to solution yet.
This thesis will have two parts. The first part is a literature study on which methods/models exist already to do transfer learning in time series applications. The second (and bigger) part will be to try to train a model that can be reused on a different application. This can either be on generated data or publicly available data.
It is unsure what the potential is of transfer learning for time series. Therefore, evaluation will not be based on how well the model works. More relevant is:
- The literature study
- The implementation and reasoning behind the different attempts that are made to get to a possibly working model.
The final objective is to create a pretrained model that can be reused for customer use cases at ML6. And, as at ML6 we believe in open-source community, a working model can be made publicly available.
2. The Intelligent Illustrator
Illustrator - Natural Language Processing - Generative Adversarial Networks - Computer Vision - Machine Learning - Artificial Intelligence
Images, illustrations or charts make a lot of things more digestible. Unfortunately, finding or creating the right illustration or image takes some time. Google images can help you with that, but sometimes does not give you the expected results. At ML6, we want to work on an intelligent illustrator that can help us generate or edit images or illustrations.
The thesis will have two parts. The first one is a literature study on what has already been done in this field. This will primarily focus on Generative Adversarial Networks (GANs). The second part is to implement these techniques yourself and to make a basic version of the “intelligent illustrator” that supports broad application.
It is unsure what the potential is of an intelligent illustrator. Therefore, evaluation will not be based on how well the model works. More relevant is:
- The literature study
- The implementation and reasoning behind the different attempts that are made to get to a possibly working model
A smart illustrator has several features:
- Generating images based on text
- Generating illustrations based on text
- Editing objects within images (bigger, smaller, other colour…)
- Transform images fully
3. Drone Imagery Pseudonymization
Drone - Computer Vision - Pseudonymization - Machine Learning - Artificial Intelligence
The GDPR introduced new regulations on how to process and handle personal data. As such, meeting the GDPR requirements is key in selling software applications to customers. The anonymization of textual data is largely accepted as something that is possible, but this is not something that is true for images. For example sometimes a face detection algorithm fails and a face is not blurred. Further, vehicles are often recognizable not just by their number plates but also by logos or other characteristics. A specific research challenge with regard to drone imagery is the fact that the camera is mobile and that real-world footage often features rapid zooming in and out. In this thesis we want to mitigate these problems.
An often-used technique today in pseudonymization is to replace identifiable elements rather than obscure them, e.g. replace all names in documents by other names. That way an observer is unable to identify names that were accidentally missed by the algorithm. Recently, this technique has been transferred to the vision space, amongst others under the name of deep privacy whereby GAN-based techniques are used to replace real faces by artificially generated ones. Optimizing this technique for differing circumstances (cameras, angles) and drone-specific aspects such as motion and zooming already present reasonable challenges. Replacing number plates and other recognizable writing or logos on cars and vans present even more research subtopics.
Research and create a Machine Learning algorithm that can pseudonymize faces and vehicles in images captured by a drone taking into account the above challenges. Technologies that can be used are Python, Tensorflow, Keras and in general the Python data science and machine learning track.
4. Drone Follow Up (Amsterdam office)
Drone - Computer Vision - Machine Learning - Artificial Intelligence
Please note: this project is executed from our location in Amsterdam
When optimizing your sports activities you would like to have a coach with you real-time to analyze your movements and give you feedback. When you are on holiday and just want that perfect picture of yourself from a different angle while you are not even aware a photo should be made. When the police is chasing a bad guy in a busy street, on foot the police can’t analyze their surroundings good enough to catch the bad guy. These are all possible use cases where Drones can help out with the assistance of computer vision. In this thesis we want to make a drone follow a person in complex environments, so you can have your own pet drone following you anywhere.
A popular algorithm nowadays in the computer vision world is YOLO (You Only Look Once), this algorithm proved to be state-of-the-art in real-time object detection. But having a good sense of an object of a person in this use case is a harder problem, doing that in real-time so a drone can follow it with limited hardware is even a harder problem. Optimizing this technique for differing circumstances (cameras, angles) and drone-specific aspects such as motion and zooming already present reasonable challenges.
Research and create a Machine Learning algorithm that can follow a person or vehicle in images captured by a drone taking into account the above challenges. Technologies that can be used are Python, Tensorflow, Keras and in general the Python data science and machine learning track.
We are a team of AI experts and the fastest growing AI company in Belgium. With offices in Ghent, Amsterdam, Berlin and London, we build and implement self learning systems across different sectors to help our clients operate more efficiently. We do this by staying on top of research, innovation and applying our expertise in practice. To find out more, please visit www.ml6.eu