Three-dimensional shape cues affect human and artificial recognition systems differently
Mikayla Cutler, Luke Baumel, Joseph Tocco, William Friebel, George K Thiruvathukal, Nicholas Baker · 2026
Read Three-dimensional shape cues affect human and artificial recognition systems differently by Mikayla Cutler, Luke Baumel, Joseph Tocco, William Friebel, George K Thiruvathukal, Nicholas Baker (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/plos-pone-0338885.txt.
Description
Mikayla Cutler, Nicholas Baker, and co-authors compared how people and computer vision systems tell objects apart when shape and surface pattern disagree. Humans lean heavily on shape, while neural networks lean on texture. Their 2026 study built a huge set of images from three-dimensional models, swapping textures onto objects and varying shading, shadows, and viewing angle.
People classified the swapped objects by shape far better than any network tested. Adding shading and shadow cues narrowed the gap by helping networks, yet the benefit landed differently: people gained most from unusual viewpoints, networks from standard ones. The authors argue that humans use such cues to grasp an object's solid structure, while networks treat them as just more surface information.
Catalog record
| CREATOR | Mikayla Cutler, Luke Baumel, Joseph Tocco, William Friebel, George K Thiruvathukal, Nicholas Baker |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50975 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |