0.25
0.5
0.75
1.25
1.5
1.75
2
Seeing People with Deep Learning
Published on Sep 13, 20156319 Views
Related categories
Chapter list
Seeing People With Deep Learning - 100:00
Seeing People With Deep Learning - 200:42
Seeing People With Deep Learning - 301:16
Seeing Humans - 101:50
Seeing Humans - 203:06
ML for Vision04:03
Challenges Lie Ahead04:54
This Lecture05:39
Pose Estimation - 106:16
Pose Estimation - 206:20
DNNs for Precise Localization? - 108:26
DNNs for Precise Localization? - 209:00
CNNs for Pose Estimation10:56
Output:Pose Confidence Maps12:31
Spatial Model13:16
Spatial priors13:52
Face prior17:12
DeepPose17:55
Cascade of pose regressors19:14
Pose Estimation Datasets20:41
MPII Human Pose21:45
Metrics22:03
State-of-the-art23:22
Tracking25:04
3-D Human Pose Tracking - 125:13
3-D Human Pose Tracking - 225:49
3-D Human Pose Tracking - 327:10
Prior Models of Human Pose and Motion27:52
Implicit Mixtures of CRBMs28:44
Bayesian Filtering w/ imCRBM31:13
Restricted Boltzmann Machines (RBM) - Review - 132:14
Restricted Boltzmann Machines (RBM) - Review - 232:37
Conditional Restricted Boltzmann Machines (CRBM) - 334:14
Conditional Restricted Boltzmann Machines (CRBM) - 134:39
Conditional Restricted Boltzmann Machines (CRBM) - 234:40
Conditional Restricted Boltzmann Machines (CRBM) - 437:18
Implicit mixture of CRBMs (imCRBM) - 137:26
Implicit mixture of CRBMs (imCRBM) - 238:03
Advantages of the imCRBM39:36
Tracking via Bayesian Filtering - 139:55
Tracking via Bayesian Filtering - 240:51
Tracking via Bayesian Filtering - 242:02
Bayesian Filtering - 142:06
Bayesian Filtering - 243:27
Experiments45:11
Multi-view: Walking + Jogging with Transitions - 146:15
Multi-view: Walking + Jogging with Transitions - 247:05
Monocular tracking with transitions48:56
Activity / Gesture49:35
Hybrid Unsupervised/Supervised52:12
Gated RBM (Two views)52:40
Convolutional Gated RBM55:00
Feature extraction examples Feature55:48
Recognition Architecture56:44
Stacked Convolutional Independent Subspace Analysis57:16
Convolution and Stacking57:56
Spatio-Temporal Feature Extraction58:30
Velocity and Orientation Selectivity59:14
Coupling of motion and invariance59:17
Motion synchrony01:00:12
Practically: how to check for synchrony?01:01:36
Learning to detect synchrony01:02:58
Results01:03:38
End-to-end Supervised01:04:40
3D Convnets for Activity Recognition01:05:07
Early CNN Architecture01:05:33
State-of-the-art CNN Architecture01:09:09
Recognizing intentional gestures01:09:56
A multi-scale architecture01:12:22
Single-scale deep architecture01:13:55
Articulated Pose: Input01:14:56
Depth Video Stream01:16:16
Training algorithm01:16:35
Initialization: structured weights01:18:04
Slightly different view01:19:20
2014 ChaLearn Looking at People Challenge (ECCV)01:20:39
Error evolution during iterative training01:21:04
Dropout (review)01:22:18
Moddrop - dropout on shared layer01:22:49
Moddrop: modality-wise dropout01:24:11
Moddrop results01:24:48
Summary01:25:14