Creating an interactive tool that accelerates the learning of annotation instructions
Examensarbete för masterexamen
Machine learning is a rapidly growing industry and plays a crucial role to make autonomous mobility safe and possible. To create these machine learning models, high-quality annotations are of great importance. One company that provides this type of data is Omara. The annotation process is a human-driven process where data is labelled based on objects of interest for that specific machine learning model. For the annotators to know what and how they should annotate the data, instructions are required. The communication of these instructions, which today is via PDF files, is the area of interest in this project. The goal of this project was to suggest how instructions can be communicated in order to accelerate the annotators' learning process, and thereby potentially increase the quality of the annotations and reduce the time needed to produce them. This was conducted by initially executing a user study where the current situation was analysed, and problems identified. This led to the creation of a list of requirements which was the foundation for the ideation and creation of concepts for a future solution. Through theoretical- and user evaluations all but one concept could be filtered out, a web-based UI that facilitates the instructions needed to annotate. The UI helps to meet requirements by, to name a few of the most important aspects, providing a consistent user experience, an easy overview of the instructions, and possibilities to easily update. This concept was then refined, and a high-fidelity prototype in Figma was created.
Annotation instructions, Communicating Instructions, User Interface (UI), User Research, Prototyping