Search Papers | Poster Sessions | All Posters
Poster B77 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Delineating the set of faces perceived as natural in the Basel Face Model
Veronica Bossio1, Wenxuan Guo1, Jasper van den Bosch2, Nikolaus Kriegeskorte1; 1Columbia University, 2University of Leeds
The Basel Face Model (BFM) provides an important tool for human face perception research, enabling the generation of realistic face images from a latent space of 3D shape and texture defining variables. The latent space is designed as an isotropic normal distribution reflecting the distribution of the 200 human faces whose 3D scans formed the basis of the BFM. However, this distribution does not reflect which of the faces look like natural human faces to people. We collected binary judgements of the naturalness of BFM faces and offer a model that predicts the probability that any BFM face will be judged as natural. This model contributes to our understanding of human face perception and will be useful to face perception researchers looking to sample natural-looking BFM faces.
Keywords: face perception generative models psychophysics