en
0.25
0.5
0.75
1.25
1.5
1.75
2
Cost-Sensitive Top-Down/Bottom-Up Inference for Multiscale Activity Recognition
Published on Nov 12, 20124062 Views
This paper addresses a new problem, that of multiscale activity recognition. Our goal is to detect and localize a wide range of activities, including individual actions and group activities, which may
Related categories
Chapter list
Cost-Sensitive Top-down/Bottom-up Multi-scale Activity Recognition00:00
Problem – Given00:11
Problem – Goal00:32
Contributions01:10
Prior Work – Punctual/Repetitive Activities01:52
Prior Work – Structured Activities02:00
Our Approach02:06
Our Approach – Related Prior Work02:52
Model: And-Or Graph (1)03:40
Model: And-Or Graph (2)03:53
Model: And-Or Graph (3)04:09
Model: And-Or Graph (4)04:40
Model: And-Or Graph (5)04:56
Inference (1)05:27
Inference (2)05:29
Inference (3)05:45
Inference: Structure05:53
Inference: α – Process06:05
Inference: β – Process06:15
Inference: ɣ – Process06:18
Inference – α, β, ɣ Processes06:49
α – Process07:03
β, ɣ – Process07:17
β – Process07:30
ɣ – Process07:37
Cost-Sensitive Inference07:43
Explore/Exploit (1)08:19
Explore/Exploit (2)08:21
Explore/Exploit (3)09:24
Explore/Exploit (4)09:38
Explore/Exploit (5)09:42
Explore/Exploit (6)09:50
Explore/Exploit (7)09:52
Explore/Exploit (8)09:52
Explore/Exploit (9)09:53
Explore/Exploit (10)09:54
Explore/Exploit (11)09:55
Explore/Exploit (12)09:56
Explore/Exploit (13)09:57
Explore/Exploit (14)09:58
Explore/Exploit (15)09:59
Explore/Exploit (16)09:59
Q-Learning10:13
Varying the Explore/Exploit trade-off10:38
Varying the Number of Detectors Run10:55
New Dataset11:20
Domain Knowledge11:42
New Dataset11:48
Available Datasets12:19
Queries Example12:42
All Parse Graphs for Group Queries13:15
All Parse Graphs for Individual Queries13:32
Results – Courtyard Dataset13:42
Conclusion14:10
Acknowledgments14:24
Questions14:29