Modeling visual memorability assessment with autoencoders reveals characteristics of memorable images
Elham Bagheri, Yalda Mohsenzadeh · 2026
Read Modeling visual memorability assessment with autoencoders reveals characteristics of memorable images by Elham Bagheri, Yalda Mohsenzadeh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13542106.txt.
Description
Elham Bagheri and Yalda Mohsenzadeh model image memory in a 2026 study. They asked what makes some images stick after one viewing. Their tool is an autoencoder. That network squeezes an image into a compact code. It then rebuilds the image from the code. Memorability is a measurable property of images.
Images the network rebuilt badly tended to be memorable. The distinctiveness of each code also tracked memorability. A small prediction network read the codes. It forecast memorability scores robustly. Heat maps highlighted the visual features behind the forecasts. The authors argue this learned space mirrors human memory. Hard-to-rebuild images linger, much as odd scenes linger for people.
Catalog record
| CREATOR | Elham Bagheri, Yalda Mohsenzadeh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72884 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |