Skip to main content
. 2023 Feb 15;14:1047694. doi: 10.3389/fpsyg.2023.1047694

Figure 3.

Figure 3

VOneNet with blur training. (A) Low-pass filtered object image recognition. (B) band-pass filtered object image recognition, (C) jumbled/occluded image recognition, and (D) shape-texture-cue-conflict image recognition. Although VOneNet has fixed V1-like first-stage mechanisms, the effects of blur training are similar to those of AlexNet. Error bars represent 95% confidence intervals estimated from the performances of eight models trained with different random seeds.