Visual Attention for Rendered 3D Shapes

1University of Lyon, CNRS, LIRIS, France 2University of Haute-Alsace, LMIA, France 3University of Strasbourg, CNRS, ICube, France 4University of Poitiers, CNRS, XLIM-SIC, France

In Computer Graphics Forum (Proceedings of Eurographics 2018)

Teaser
We conducted eye tracking experiments on rendered 3D objects. The human saliency information is mapped on the 3D meshes (in the form of fixation density maps) and serves to study which factors influence human attention and to evaluate state-of-the-art saliency algorithms. At the bottom we show the Pearson correlation between saliency maps from humans and algorithms

Abstract

Understanding the attentional behavior of the human visual system when visualizing a rendered 3D shape is of great importance for many computer graphics applications. Eye tracking remains the only solution to explore this complex cognitive mechanism. Unfortunately, despite the large number of studies dedicated to images and videos, only a few eye tracking experiments have been conducted using 3D shapes. Thus, potential factors that may influence the human gaze in the specific setting of 3D, are still to be understood. In this work, we conduct two eye-tracking experiments involving 3D shapes, with both static and time-varying camera positions. We propose a method for mapping eye fixations (i.e., where humans gaze) onto the 3D shapes with the aim to produce a benchmark of 3D meshes with fixation density maps, which is publicly available. First, the collected data is used to study the influence of shape, camera position, material and illumination on visual attention. We find that material and lighting have a significant influence on attention, as well as the camera path in the case of dynamic scenes. Then, we compare the performance of four representative state-of-the-art mesh saliency models in predicting ground-truth fixations using two different metrics. We show that, even combined with a center-bias model, the performance of 3D saliency algorithms remains poor at predicting human fixations. To explain their weaknesses, we provide a qualitative analysis of the main factors that attract human attention. We finally provide a quantitative comparison of human-eye fixations and Schelling points and show that their correlation is weak.

Paper (16MB)
Supplementary Material (37MB)

Dataset

All data can be downloaded via the following links
Original models (48MB) - 32 full resolution 3D shapes (OBJ format), originaly used to acquire the fixation maps and compute the saliency from automatic algorithms.
Simplified models (13MB) - 32 shapes with decreased resolution (OBJ format), used to compute the statistics.
Fixation maps (2MB) - ASCII files with per-vertex fixation values. 96 fixation files, one for each view of each shape.
Saliency maps (27MB) - ASCII files with per-vertex saliency values. 384 saliency files, one for each shape, algorithm (Lee, Leifman, Song and Tasse) and  blurring factor (3).
Centricity and visibility maps (32MB) - ASCII files with per-vertex visibility values (96 files) and ASCII files with per-vertex centricity values (for 13 different Gaussian radius) (1248 files).

Acknowledgements

We are deeply grateful to Flora Ponjou Tasse, George Leifman, Ran Song and Chang Ha Lee for kindly running their saliency algorithms on our dataset (or providing their source code).