StatTexNet: Evaluating the Importance of Statistical Parameters for Pyramid-Based Texture and Peripheral Vision Models
Submission Type: Full Paper
Keywords: peripheral vision, texture synthesis, multi-scale pyramid, statistic selection, contrastive learning
TL;DR: Finding important statistics for pyramid-based peripheral vision models through self-supervised learning.
Abstract: Peripheral vision plays an important role in human vision, directing where and when to
make saccades. Although human behavior in the periphery is well-predicted by pyramid-
based texture models, these approaches rely on hand-picked image statistics that are still
insufficient to capture a wide variety of textures. To develop a more principled approach to
statistic selection for texture-based models of peripheral vision, we develop a self-supervised
machine learning model to determine what set of statistics are most important for repre-
senting texture. Our model, which we call StatTexNet, uses contrastive learning to take a
large set of statistics and compress them to a smaller set that best represents texture fami-
lies. We validate our method using depleted texture images where the constituent statistics
are already known. We then use StatTexNet to determine the most and least important
statistics for natural (non-depleted) texture images using weight interpretability metrics,
finding these to be consistent with previous psychophysical studies. Finally, we demonstrate
that textures are most effectively synthesized with the statistics identified as important;
we see noticeable deterioration when excluding the most important statistics, but minimal
effects when excluding least important. Overall, we develop a machine learning method of
selecting statistics that can be used to create better peripheral vision models. With these
better models, we can more effectively understand the effects of peripheral vision in human
gaze.
Supplementary Material: zip
Submission Number: 21
Loading