The Image Explorer
Short description
In May, we will begin designing and developing a prototype for image exploration within the framework of GPS400. The goal is to create new and innovative ways of interacting with the centre’s extensive image collections. With the support of AI, we will explore how machine-based interpretations of images can enable novel ways of navigating, comparing, contrasting, and grouping visual material – offering fresh perspectives and new understandings. The project is a collaboration between Appademin at the Department of Applied IT and the steering and research group of GPS400.
What happens when we use AI to help us look at images in new ways? In an exploratory project, AI models are being used to identify hidden connections between images. The aim is not to make searches more efficient, but to open up new possibilities for interpretation and understanding. Rather than knowing in advance what we are looking for, we allow the technology to reveal unexpected relationships.
The AI models used in the project can compare and connect images based on how they are described and interpreted, rather than solely on what they depict.
This makes it possible to identify nuances and themes that may not be apparent at first glance. It could contribute to new ways of analysing historical archives, artistic projects and large image collections. The aim is not simply to find the right image, but to allow unexpected connections to generate new forms of understanding.
Interpreting and Navigating a New Visual Landscape
Making use of the new possibilities offered by the technology requires an active and media-literate approach. Interpreting the patterns and connections suggested by the model becomes a central part of the process.
This is where technological innovation meets visual analysis. Researchers and users need to draw on their experience and critical judgement to make sense of what the AI system reveals. If the prototype currently under development proves successful, the ambition is to develop it further and make the tool available to a wider audience, subject to future funding.
How does the model work?
The project uses pre-trained AI models
No custom training is required
Unlike earlier AI projects, which required millions of images to train models from scratch, this project uses models that have already been trained.
Focus on relationships, not just content
The model identifies similarities and differences based on how images are interpreted and described, rather than simply on what they depict.
Navigation rather than search
The aim is not to find a specific image, but to enable users to explore unexpected connections and develop new understandings through them.