Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session C: Friday, August 15, 2:00 – 5:00 pm, de Brug & E‑Hall

Memory diffusion: Using generative AI to create an image database for memory research

Fabian Kamp¹, Frieda Josefine Born², Bernhard Spitzer³; ¹Max Planck Schools, ²Technische Universität Berlin, ³Max-Planck Institute

Presenter: Fabian Kamp

When we memorize visual stimuli, their content is processed at multiple levels, ranging from the fine-grained perceptual details to the semantic concepts and categories. However, it is unclear to which extend low- and high-level information is maintained in memory over time. Real-world stimuli are not ideal for investigating this question, as they often exhibit strong correlations between processing levels: Conceptually similar objects tend to share similar visual features. Using generative AI we created a new database of 496 image pairs orthogonalizing semantic (word2vec) and perceptual (CoreNet-S) information. Specifically, we generated image pairs that either (a) depict objects from distinct semantic concepts but are perceptually similar, or (b) show the same object but are perceptually dissimilar.

Topic Area: Object Recognition & Visual Attention

Extended Abstract: Full Text PDF