en
0.25
0.5
0.75
1.25
1.5
1.75
2
People Watching: Human Actions as a Cue for Single View Geometry
Published on 2012-11-125949 Views
We present an approach which exploits the coupling between human actions and scene geometry. We investigate the use of human pose as a cue for single-view 3D scene understanding. Our method builds up
Related categories
Presentation
People Watching: Human Actions as a Cue for Single-View Geometry00:00
Indoor Single-View 3D Geometry00:03
Where are the people?00:39
But People Are Interesting!00:58
People as Clutter?01:07
People – Cues not Clutter01:28
Affordances – Where can I Sit?01:51
Our Goal – Inverse Problem02:33
Our Approach (1)02:48
Our Approach (2)03:01
Our Approach (3)03:05
Our Approach (4)03:17
Our Approach (5)03:24
Our Approach (6)03:29
Our Approach (7)03:42
Detecting Human Actions (1)03:48
Detecting Human Actions (2)04:10
Additional Detectors04:27
DPM Detections04:35
Our Approach (8)04:46
From Poses to Functional Regions (1)04:55
From Poses to Functional Regions (2)05:32
Affordance Constraints05:44
Our Approach (9)05:51
3D Room Hypotheses06:01
Appearances06:19
Appearances Can Be Deceiving06:39
What If We Observe People?06:57
Penalties From Functional Regions (1)07:04
Penalties From Functional Regions (2)07:41
Reranking Results07:57
Our Approach (10)08:03
Estimating free space (1)08:09
Estimating free space (2)08:28
Estimating free space (3)08:36
Results08:41
Qualitative Example (1)08:47
Qualitative Example (2)09:25
Quantitative Results - Timelapses09:52
Single Images with People (1)10:18
Single Images with People (2)10:37
Quantitative Results – Single Image10:49
Limitations11:07
Our Work on Semantics11:38
Conclusions12:17
Thank You12:50