Master Thesis - Language Models for Configurable Sensors and Programming Libraries

Vision libraries and configurable software can contain hundreds or thousands of tools and functions for analyzing and processing images. Although the user is aware of what problem they want so solve, it can be very difficult to know which tools are available or which ones to use. This poses an unwanted learning curve where even a vision expert can struggle to solve their application.

The topic of this thesis will be to abstract the algorithms and tools from the user and instead focus on the problems that they want to solve. This will make it easier for them to find what they want instead of reading and searching in manuals.

Example research questions:

• Can we use a pre-trained open-sourced language model and fine-tune it?
• What type of data do we need to provide? Is an existing user manual enough?
• What level of "abstraction" can we reach?

"Show me functions that operate on binary images"

"I want to measure the dimensions of my product"

Prerequisites

The applicant will benefit from experience in the following fields:

• Interest in Natural Language Processing
• Interest in UX
• Python or similar scripting languages
• Deep learning framework such as TensorFlow

Contact

For more information, contact R&D Manager Andreas Wrangsjö, andreas.wrangsjo@sick.se

We look forward to your application before 18th of October!