0.25
0.5
0.75
1.25
1.5
1.75
2
Domain Adaptation for Upper Body Pose Tracking in Signed TV Broadcasts
Published on Apr 03, 20142113 Views
The objective of this work is to estimate upper body pose for signers in TV broadcasts. Given suitable training data, the pose is estimated using a random forest body joint detector. However, obtain
Related categories
Chapter list
Domain Adapta,on for Upper Body Pose Tracking in Signed TV Broadcasts00:00
Motivation: Automatic sign language recognition00:00
Introduction00:35
Problem - 101:07
Problem - 201:27
Problem - 301:42
One solution - 101:57
One solution - 202:06
One solution - 302:11
Key idea in this paper - 102:34
Key idea in this paper - 202:53
Example: Tracking signers wearing short sleeves - 103:08
Example: Tracking signers wearing short sleeves - 203:35
Related work03:50
Method04:18
Existing method for signers with long sleeves - 104:43
Existing method for signers with long sleeves - 205:25
Existing method for signers with long sleeves - 305:39
Existing method for signers with long sleeves - 405:50
Solution06:30
Method overview - 106:52
Method overview - 207:21
Method overview - 307:32
Method overview - 407:44
Method detail - 108:06
Method detail - 208:30
Method detail - 308:53
Method detail - 408:58
Method detail - 509:29
Method detail - 609:30
Method detail - 709:39
Method : Template personalisation - 110:01
Method : Template personalisation - 210:19
Method : Template personalisation - 310:22
Method : Template personalisation - 411:01
Method : Template personalisation - 511:24
Method : Template personalisation - 611:43
Method : Template personalisation - 711:53
Method detail - 812:11
Method detail - 912:20
Method detail - 1012:26
Method : example output12:31
Results12:57
Dataset - 113:04
Dataset - 213:21
Quantitative results - 113:37
Quantitative results - 214:01
Quantitative results - 314:42
Conclusion15:02