Joint Image-EEG Embedding (Master Computer Science Project, Ongoing)
Author
Description
Multimodal multimedia retrieval approaches, which extract feature vectors into a co-embedding space shared between multimedia modalities, have recently gained prominence as they allow searching not only within a single modality domain but also between domains. Common multimedia domains include text, images, audio, video, and 3D models. However, the principle of co-embedding is not limited to these domains; any digital format that has a correlation in terms of content should, in principle, be able to be transformed into a common embedding space. In this project, we explore whether Electroencephalography (EEG) signals captured while viewing an image can be embedded in an image-EEG co-space. Furthermore, we want to investigate whether the vectors extracted from the EEG can be used to search for the correlated images and whether the results differ significantly from a random result. The goal of the project is to determine whether the technique is feasible in principle.
Start / End Dates
2024/09/13 - 2025/02/14