Automatic Tagging Suggestion for Database Enrichment
Media organisations often use images from archives, for example to accompany news articles. By developing a tool for automatic image tagging, we intend to help organisations in ensuring the consistency and accuracy of the tags. This should ease image retrieval.
Objective
We aim to determine the feasibility of creating an automatic image tagging tool capable of effectively handling a growing number of tags. This is important in the context of news and journalism, where new tags need to be added regularly.
Results
The project will seek to address the following research questions:
- How does the accuracy of the automatic image tagging tool compare to that of a tool based on a more traditional image classifier?
- Is the automatic image tagging tool able to maintain its performance as the number of tags increases?
- What is the minimum number of images with a new tag required for the tool to produce acceptable results?
Duration
01 March 2023 - 01 July 2023
Approach
The initial stage of the project involves designing a tool that adheres to the specified requirements. Once the tool is constructed and operational, the second phase will involve evaluating its performance as new tags are introduced and associated with a set of images.
HU researchers involved in the research
Collaboration with knowledge partners
NOS: a public broadcasting company that develops radio and television shows about news, events and sports.
