A simple smartphone-based stereoscope
Troubleshooting:
aA
-> Hide Toolbar
in the address barThese images were invented by scientists to understand how stereo vision works in humans. The sense of depth is only apparent when they are viewed stereoscopically. In many cases, the perceived shape is strongly influenced by "unmatched features", meaning features that are visible to one eye but not the other. Beside each thumbnail is a diagram of the perceived shape relative to two eyes, with the "unmatched features" drawn in red.
These are interactive versions, where the stimulus is meant to change as you rotate your head horizontally.
View your own images stereoscopically by pasting two URLs and clicking Go.
(Higher resolution images produce a better effect but take longer to load. If your
screen is persistently blank in VR mode, check the image server's access controls.)
Another way to see your own images is by passing their URLs as query strings. The syntax is
(base URL)?lImg=(left image URL)&rImg=(right image URL)
For example, an equivalent way to launch the Two Planes stimulus above is by typing
StereoPages/static.html?lImg=/stereoscope/assets/StereoImages/1_1/2048one_l.png
&rImg=/stereoscope/assets/StereoImages/1_1/2048one_r.png
This VR Stereoscope is built in A-Frame, with contributions from
Oliver Habert, Anubha Srivastava, and Jialiang Wang.
The code is minimal. You can
Fork
on github or
The perceptual stimuli were collected in the following papers to study perceptually-consistent computer vision stereo algorithms. Our algorithms properly use the occlusion cue, and as a result, they work in most of these stimuli whereas many state-of-the-art algorithms fail completely. However, the Wallpapers, Cross, Folded Paper and Cylinders stimuli remain open research problems.
For more, including supplementary materials, posters, data and slides, see this project page.
Toward perceptually-consistent stereo: A scanline study
Jialiang Wang, Daniel Glasner, and Todd Zickler
In Proc. International Conference on Computer Vision (ICCV), 2017.
Local Detection of Stereo Occlusion Boundaries
Jialiang Wang and Todd Zickler
In Proc. Computer Vision and Pattern Recognition (CVPR), 2019.