Download presentation
Presentation is loading. Please wait.
Published byXavier O'Leary Modified over 10 years ago
1
1 Hierarchical Part-Based Human Body Pose Estimation * Ramanan Navaratnam * Arasanathan Thayananthan Prof. Phil Torr * Prof. Roberto Cipolla * University Of Cambridge Oxford Brookes University
2
2Introduction Input
3
3Introduction Output
4
4Overview 1.Motivation 2.Hierarchical parts 3.Template search 4.Pose estimation in a single frame 5.Temporal smoothing 6.Summary & Future work
5
5Overview 1.Problem motivation ??? 2.Hierarchical parts 3.Template search 4.Pose estimation in a single frame 5.Temporal smoothing 6.Summary & Future work
6
6Overview 1.Problem motivation ??? 2.Hierarchical parts 3.Template search 4.Pose estimation in a single frame 5.Temporal smoothing 6.Summary & Future work
7
7Overview 1.Problem motivation ??? 2.Hierarchical parts 3.Template search 4.Pose estimation in a single frame 5.Temporal smoothing 6.Summary & Future work
8
8Motivation Real-time Object Detection for Smart Vehicles – D. M. Gavrila & V. Philomin (ICCV 1999) Filtering using a tree-based estimator – Stenger et.al. (ICCV 2003)
9
9Motivation Exponential increase of templates with dimensions Real-time Object Detection for Smart Vehicles – D. M. Gavrila & V. Philomin (ICCV 1999) Filtering using a tree-based estimator – Stenger et.al. (ICCV 2003)
10
10Motivation Pictorial Structures for Object Recognition – P. Felzenszwalb & D. Huttenlocher (IJCV 2005) Human upper body pose estimation in static images – M.W. Lee & I. Cohen (ECCV 2004)
11
11Motivation Part based approach Assembling parts together is complex Pictorial Structures for Object Recognition – P. Felzenszwalb & D. Huttenlocher (IJCV 2005) Human upper body pose estimation in static images – M.W. Lee & I. Cohen (ECCV 2004)
12
12Motivation Automatic Annotation of Everyday Movements – D. Ramanan & D. A. Forsyth (NIPS 2003) 3-D model-based tracking of humans in action:a multi-view approach – D. M. Gavrila & L. S. Davis (CVPR 1996)
13
13Motivation Automatic Annotation of Everyday Movements – D. Ramanan & D. A. Forsyth (NIPS 2003) 3-D model-based tracking of humans in action:a multi-view approach – D. M. Gavrila & L. S. Davis (CVPR 1996) State space decomposition
14
14 Hierarchical Parts
15
15 Hierarchical Parts
16
16 Hierarchical Parts
17
17 Hierarchical Parts
18
18 Hierarchical Parts Conditional prior p(x i /x parent(i) ) Spatial dimensions (translation) Joint Angles
19
19 Hierarchical Parts Head and torso Upper arm Lower Arm False Positive True Positive
20
20 Hierarchical Parts Detection Threshold = 0.81 Detections Head and torso 6156 Part
21
21 Hierarchical Parts Detection Threshold = 0.81 Detections Head and torso 6156 13 19944 993 Part Lower arm
22
22 Template Search
23
23 Template Search
24
24 Template Search
25
25 Template Search Features Chamfer distance Appearance
26
26 Template Search Features Chamfer distance Appearance
27
27 Template Search Features Chamfer distance Appearance
28
28 Template Search Features Chamfer distance Appearance
29
29 Template Search Features Chamfer distance Appearance
30
30 Template Search Features Chamfer distance Appearance
31
31 Template Search Features Chamfer distance Appearance
32
32 Template Search Features Chamfer distance Appearance
33
33 Template Search Features Chamfer distance Appearance
34
34 Template Search Learning Appearance Match T pose based on edge likelihood only in initial frames Update 3D histograms in RGB space that approximates P(RGB/part) and P(RGB)
35
35 Pose Estimation in a Single Frame
36
36 Pose Estimation in a Single Frame
37
37 Pose Estimation in a Single Frame
38
38 Temporal Smoothing HMM
39
39 Temporal Smoothing HMM T = t
40
40 Temporal Smoothing HMM Viterbi back tracking
41
41 Temporal Smoothing Viterbi back tracking
42
42 Temporal Smoothing
43
43 Summary & Future work Summary Realtime process (unoptimized code at 1Hz, 2.4 Ghz IG RAM) 3D pose Automatic initialisation and recovery from failure
44
44 Summary & Future work Summary Realtime process (unoptimized code at 1Hz, 2.4 Ghz IG RAM) 3D pose Automatic initialisation and recovery from failure Future work Extend robustness to illumination changes Non-fronto-parallel poses Poses when arms are inside the body silhouette Simple gesture recognition by assigning semantics to regions of articulation space
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.