Search Papers | Poster Sessions | All Posters
Poster C8 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Predicting human perceptual similarity and memory false alarms using visual and semantic deep neural networks
Natalia Kurilenko1 (), Kevin J. M. Le2, Juri Minxha2, Ueli Rutishauser1; 1Cedars-Sinai Medical Center, 2California Institute of Technology
Despite considerable effort, predicting human similarity judgments and aspects of memory that rely on such judgments remains challenging. In this work, we collected a large set of human similarity judgments and compared the performance of semantic and visual deep neural networks in predicting them. We then examine the effectiveness of the computational similarity metrics in predicting false alarms in a recognition memory task. We show that general visual features best predict perceptual similarity while combined visual and semantic information better explain memory performance.
Keywords: perceptual similarity deep learning visual recognition memory