Integrating Computational and Neural Findings in Visual Object Perception

78.6K
views
33
authors
11
articles
Editors
3
Impact
Loading...
Overview of the voxel-wise modeling (VM) procedure used in this study. (A) Human subjects were shown 1260 natural images while (B) fMRI data were recorded. (C–E) These data were modeled as a function of three different feature spaces. Each feature space reflects a different hypothesis about which features are represented in scene-selective areas. (C) For the Fourier power mode, the feature space was computed by taking the Fourier transform of each stimulus image and then averaging the amplitude spectrum over the orientation and spatial frequency bins shown at right. (D) For the subjective distance model, the feature space consisted of ratings from three humans who judged whether the main content of each stimulus scene was (1) < 2 ft away, (2) < 4 ft away, (3) < 20 ft away, (4) < 100 ft away, and (5) >100 ft away. (E) For the semantic category model the feature space consisted of labels from three human raters who labeled the objects in each stimulus image using 19 semantic labels. (F) Ordinary least squares regression was used to find a set of weights (β) that map the features in each model onto the BOLD responses in each voxel. Each feature space and its associated β weights constitute a different encoding model. (G) In order to validate the models in an independent data set, the same subjects were shown a different set of 126 images while (H) fMRI responses were collected. (I) To assess model accuracy, the β weights estimated from the training data were used to predict responses in this withheld model validation data set. (J) To reveal patterns of tuning in the features quantified by each different model, pre-specified t contrasts were computed between β weights in each model and projected onto the cortical surface, and β weights were averaged over voxels in different regions of interest and plotted.
10,119 views
113 citations

Modeling human aesthetic perception of visual textures is important and valuable in numerous industrial domains, such as product design, architectural design, and decoration. Based on results from a semantic differential rating experiment, we modeled the relationship between low-level basic texture features and aesthetic properties involved in human aesthetic texture perception. First, we compute basic texture features from textural images using four classical methods. These features are neutral, objective, and independent of the socio-cultural context of the visual textures. Then, we conduct a semantic differential rating experiment to collect from evaluators their aesthetic perceptions of selected textural stimuli. In semantic differential rating experiment, eights pairs of aesthetic properties are chosen, which are strongly related to the socio-cultural context of the selected textures and to human emotions. They are easily understood and connected to everyday life. We propose a hierarchical feed-forward layer model of aesthetic texture perception and assign 8 pairs of aesthetic properties to different layers. Finally, we describe the generation of multiple linear and non-linear regression models for aesthetic prediction by taking dimensionality-reduced texture features and aesthetic properties of visual textures as dependent and independent variables, respectively. Our experimental results indicate that the relationships between each layer and its neighbors in the hierarchical feed-forward layer model of aesthetic texture perception can be fitted well by linear functions, and the models thus generated can successfully bridge the gap between computational texture features and aesthetic texture properties.

15,174 views
25 citations