Internship | Hierarchical object detection with coarse-to-fine labels and uncertainty

Introduction
Humans tend to label objects with fine-grained names when they are an expert (e.g. sea eagle) and more coarsely if they are uncertain (e.g. bird). We want a neural network to do the same, leveraging available knowledge about hierarchies (living creature, animal, bird, predator bird, eagle, sea eagle). Some promising directions have been proposed. One approach is to use the hierarchy within the loss function: the more steps to go from actual label to predicted label, the higher the loss [1]. E.g. in the case of an eagle, the loss for duck is smaller than snake. A second approach is the transferable belief model, which represents uncertainty by distributing the probabilities over the powerset of all options [2], which could be applied to a hierarchy. E.g. it is not certain whether it is a eagle or duck, but it is completely certain that it is a bird. This could be use into a loss function. A third approach is hyperbolic learning [3]. It uses the hierarchy to create embeddings for the super/sub classes in hyperbolic space. This can be applied to many computer vision tasks [3], and recently proved useful for object detection [3, 4], with the advantage of indicating uncertainty [5]. In hyperbolic learning, the intermediate classes serve as implicit guidance, via the embedding space. We are curious whether more explicit guidance, like in [1] and [2], can be combined with the strengths of hyperbolic learning [3-5], for multi-level supervision?

[1] Hierarchical embeddings
[2] Transferable belief model
[3] Hyperbolic deep learning in computer vision
[4] Hierarchical, hyperbolic object detection
[5] Hyperbolic object and scene representation
[6] Hyperbolic semantic uncertainty

What you will be doing
Dive into the world of hyperbolic space, class hierarchies, uncertainty modelling, hierarchy-based loss functions and explore the opportunities of hierarchical learning for object detection. Research how models can learn to classify from coarse to fine. Investigate how for unknown fine-grained labels there can still be good coarse-level predictions.

You will perform this assignment within TNO’s Intelligent Imaging department. The Intelligent Imaging department is a passionate, creative, and dedicated team of professionals (60 people) specialized in developing groundbreaking applications in the field of computer vision. Our team members have diverse backgrounds, ranging from the medical field to artificial intelligence. Intelligent Imaging is a young and growing department which has built up a lot of expertise over the past years in AI and deep learning.

We are looking for a student that wants to join our cutting-edge research team to explore the opportunities of hierarchical learning for object detection. This position is perfect for students who are passionate about AI and advanced machine learning techniques.

Additional requirements include that you are in the final stages of your master's degree in artificial intelligence, computer science, physics, mathematics, electrical engineering, or a similar degree. You have some experience in computer vision, artificial intelligence, deep learning and Python-programming. The duration of the master project or internship is between 6 and 12 months.

You want an internship opportunity on the precursor of your career; an internship gives you an opportunity to take a good look at your prospective future employer. TNO goes a step further. It’s not just looking that interests us; you and your knowledge are essential to our innovation. That’s why we attach a great deal of value to your personal and professional development. You will, of course, be properly supervised during your work placement and be given the scope for you to get the best out of yourself. Furthermore, we provide:

• A highly professional, innovative internship environment, within a team of top experts.
• A suitable internship allowance (615 euro for wo-, hbo- and mbo-students, for a full-time internship).
• Possibility of eight hours of free leave per internship month (for a full-time internship).
• A free membership of Jong TNO, where you can meet other TNO professionals and join several activities, such as sports activities, (work-related) courses or the yearly ski-trip.
• Use of a laptop.
• An allowance for travel expenses in case you don’t receive an OV-card.

At TNO, we innovate for a healthier, safer and more sustainable life. And for a strong economy. Since 1932, we have been making knowledge and technology available for the common good. We find each other in wonder and ingenuity. We are driven to push boundaries. There is all the space and support for your talent and ambition. You work with people who will challenge you: who inspire you and want to learn from you. Our state-of-the-art facilities are there to realize your vision. What you do at TNO matters: impact makes the difference. Because with every innovation you contribute to tomorrow’s world. Read more about TNO as an employer.

At TNO we encourage an inclusive work environment, where you can be yourself. Whatever your story and whatever unique qualities you bring to the table. It is by combining our unique strengths and perspectives that we are able to develop innovations that make a real difference in society. Want to know more? Read what steps we are taking in the area of diversity and inclusion.

After the first CV selection, the application process will be conducted by the concerning department. TNO will provide a suitable internship agreement. If you have any questions about this vacancy, you can contact the contact person mentioned below.

For this internship vacancy it is required that the AIVD issues a security clearance (VGB) after conducting a security screening. Take into account that this process may take about 8 weeks. If you have been abroad for more than 6 consecutive months, or if you do not have the Dutch nationality, it may take longer. Read more about security screening on the AIVD website.

Important to be aware of before applying:

• Before the start of the internship, the internship agreement from TNO needs to be signed. For students at a college or university based in the Netherlands, TNO uses the UNL-template (supplemented with a number of specific agreements from TNO). For students of foreign and MBO educational institutions, the TNO internship agreement applies. TNO does not sign any other internship agreements.
• Before the start of the internship, the educational institution will need to confirm in writing that:
  • 1) you are enrolled at the educational institution during the internship, and;
  • 2) the internship takes place as part of the programme of the study.
• The confirmation of educational institution takes place by signing the UNL template or forms prepared by TNO.
• Interns at TNO must be in possession of a Dutch residential address at the start of the internship. Performance of internship activities from abroad is not possible.