Thesis opportunity- Self-Supervised Training for Industrial Applications

What the Master Thesis is about/background to the problem to investigate

In many industrial applications deep learning solutions are sought after, but usually these applications come with some restrictions. For example:

• The data is often difficult to annotate or hard to obtain.
• Customers don’t want to place sensitive data in data centers.
• The domain the image is captured in is starkly different to common open-source datasets or other natural images.

Still there is a need for deep learning in these environments to overcome complexity of traditional image processing or to allow more people to configure solutions.

Can we overcome the performance of a network pre-trained traditionally on open-source datasets by leveraging self-supervised pre-training on the application images without annotations? Can we improve performance by incorporating images from the same domain, but outside the actual application scope?

Concretely we’d want the networks trained in this way to perform better on one or more downstream tasks. Such as, transfer learning to the actual small, annotated dataset, or as an all-purpose feature extractor for various tasks.

The master thesis work focuses on the following/example of research questions

• Which method self-supervision perform the best in an industrial context?
• Which self-supervision method is suitable for convolutional nets versus transformers?
• How many domain images without annotations are needed?
• How many annotated application images are needed?
• Is it required to use a large base dataset to reach optimum performance?

Prerequisites

An interest and solid foundation in image processing and deep learning is beneficial to assess and collect methods that are relevant to investigate. Experience in self-supervision or training of deep neural networks in general is also welcomed. Some experience in python programming is required.

To analyze and make the correct conclusions a good basis in mathematics and statistics is needed.

Contact

For more information about the position, contact:

Daniel Rydström, Group Manager Vision Algorithms, daniel.rydstrom@sick.se

or

Sarah Lantz, HR Business Partner, +46 739 10 99 37.

Welcome with your application 20^th of October at the latest!

SICK is a world-leading supplier of sensors and sensor solutions for industrial applications. We are 12 000 employees in 50 countries and our headquarter is located in Freiburg, Germany. SICK in Linköping is an innovation center for Machine Vision and we are 90 committed employees with a big interest in image processing and visualization. For more than 35 years, our team at SICK Linköping has successfully developed and delivered software for technically leading products within the field of 2D and 3D vision, as well as system solutions for i.e. robot guidance and quality control.

At SICK in Linköping, we are very proud of being a healthy and attractive workplace. For many years, we have been elected as one of the best workplaces in Sweden according to the survey Great Place to Work, the latest award is from 2023. We work actively to reduce our climate footprint and we are active in various ways to contribute to the society and to increase diversity at our workplace.