en
0.25
0.5
0.75
1.25
1.5
1.75
2
Detecting Actions, Poses, and Objects with Relational Phraselets
Published on Nov 12, 20125066 Views
We present a novel approach to modeling human pose, together with interacting objects, based on compositional models of local visual interactions and their relations. Skeleton models, while flexible
Related categories
Chapter list
Detecting Actions, Poses, and Objects with Relational Phraselets00:00
Chaitanya Desai00:06
K-way action classification - 100:11
K-way action classification - 200:29
What’s wrong with K-way classification? - 100:37
What’s wrong with K-way classification? - 200:45
Our goal: detailed action understanding - 101:03
Our goal: detailed action understanding - 201:15
Our goal: detailed action understanding - 301:20
Our goal: detailed action understanding - 401:24
Challenge 1: human pose estimation01:32
Challenge 2: person-object occlusions01:44
(Revised) action understanding01:53
Related work: PASCAL Action Classification Challenge02:21
Our approach02:42
Articulated pose estimation - 102:53
Articulated pose estimation - 203:16
Articulated pose estimation - 303:39
Visual Phrases - 103:51
Visual Phrases - 204:12
Geometric parts (poselets) - 104:22
Geometric parts (poselets) - 204:53
Approach05:08
Articulated models + visual composites05:13
Articulated models + visual composites + geometric parts05:24
Learning phraselets05:53
Clusters06:20
Model occlusions with separate clusters06:32
Local mixtures of phraselets - 106:52
Local mixtures of phraselets - 207:06
Local mixtures of phraselets - 307:22
Geometry-dependent parts07:46
Inference & Learning08:14
Possible Criticisms - 108:44
Possible Criticisms - 209:00
Possible Criticisms - 309:16
Possible Criticisms - 409:24
Experimental Results09:45
High-confidence detections10:04
High-scoring false-positives10:25
Evaluation10:54
Action classification11:00
Articulated pose estimation11:15
Action detection/ localization11:32
Action detection11:35
A look back - 112:13
A look back - 212:30
Thank you12:56