|
|
|
|
|
by wesleyy
3895 days ago
|
|
I'm still not convinced. AR through visual SLAM is not extremely difficult. All the demos here are indoor and from the occlusion handling for the first scene (0:13) you can see a slight shadow on the right of the table leg so I assume it's using some kind of structured light to sense the depth (versus multi camera/stereo systems). Outdoor tracking with this kind of approach is very difficult because sunlight basically destroys all the information. The difficult part is where the system portrays this information to the user, which they have not shown yet. Their patents indicate some kind of projection directly onto the eye but there are obvious problems with that, such as when you move your eyeball. Also, running visual slam on a mobile device at real time with a respectable framerate is another challenge. |
|
The latency on this demo looks pretty bad though.