Download presentation
Presentation is loading. Please wait.
Published byHarold Snow Modified over 9 years ago
1
Contextual models for object detection using boosted random fields by Antonio Torralba, Kevin P. Murphy and William T. Freeman
2
Quick Introduction What is this? Now can you tell?
3
Belief Propagation (BP) Network (Pairwise Markov Random Fields) observed nodes ( y i )
4
Belief Propagation (BP) Network (Pairwise Markov Random Fields) observed nodes ( y i ) hidden nodes ( x i )
5
Belief Propagation (BP) Network (Pairwise Markov Random Fields) observed nodes ( y i ) hidden nodes ( x i ) Statistical dependency, called local evidence: Shord-hand
6
Belief Propagation (BP) Statistical dependency: Local evidence Shord-hand Statistical dependency: Compatibility function
7
Belief Propagation (BP) Joint probability
8
Belief Propagation (BP) Joint probability x x1x1 x2x2 xixi …. x5x5 x3x3 x1x1 x4x4 xjxj x 12 y1y1 y2y2 yiyi
9
Belief Propagation (BP) Joint probability x x1x1 x2x2 xixi …. x5x5 x3x3 x1x1 x4x4 xjxj x 12 y1y1 y2y2 yiyi
10
Belief Propagation (BP) The belief b at a node i is represented by the local evidence of the node all the messages coming in from neighbors xixi xjxj ∏ NiNi yiyi
11
Belief Propagation (BP) The belief b at a node i is represented by the local evidence of the node all the messages coming in from neighbors xixi xjxj ∏ NiNi yiyi
12
Belief Propagation (BP) Messages m between hidden nodes How likely node j thinks it is that node i will be in the corresponding state. xixi xjxj m ji (x i )
13
Belief Propagation (BP) xixi xjxj xkxk xixi xjxj m ji (x i )
14
Conditional Random Field Distribution of the form:
15
Conditional Random Field Distribution of the form:
16
Boosted Random Field Basic Idea: Use BP to estimate P(x|y) Use boosting to maximize Log Likelihood of each node wrt to
17
Algorithm: BP Minimize negative log likelihood of training data ( y i ). Label Loss function to minimize:
18
Algorithm: BP Minimize negative log likelihood of training data ( y i ). Label Loss function to minimize:
19
Algorithm: BP Minimize negative log likelihood of training data ( y i ). Label Loss function to minimize:
20
Algorithm: BP xixi xjxj NiNi ∏ yiyi
21
xixi xjxj NiNi ∏ yiyi
22
xixi xjxj NiNi ∏
23
xixi xjxj
24
xixi F : a function of the input data yiyi
25
Algorithm: BP xixi xjxj with yiyi
26
Algorithm: BP xixi xjxj with yiyi
27
Function F xixi yiyi Boosting! f is the weak learner: weighted decision stumps.
28
Minimization of loss L
30
where
31
Local Evidence: algorithm For t=1..T Iterate N boost times find the best basis function h update local evidence with update the beliefs update the weights Iterate N BP times update messages update the beliefs xixi xjxj yiyi
32
Local Evidence: algorithm For t=1..T Iterate N boost times find the best basis function h update local evidence with update the beliefs update the weights Iterate N BP times update messages update the beliefs xixi xjxj yiyi
33
Local Evidence: algorithm For t=1..T Iterate N boost times find the best basis function h update local evidence with update the beliefs update the weights Iterate N BP times update messages update the beliefs xixi xjxj yiyi
34
Local Evidence: algorithm For t=1..T Iterate N boost times find the best basis function h update local evidence with update the beliefs update the weights Iterate N BP times update messages update the beliefs xixi xjxj yiyi
35
Local Evidence: algorithm For t=1..T Iterate N boost times find the best basis function h update local evidence with update the beliefs update the weights Iterate N BP times update messages update the beliefs xixi xjxj yiyi
36
Local Evidence: algorithm For t=1..T Iterate N boost times find the best basis function h update local evidence with update the beliefs update the weights Iterate N BP times update messages update the beliefs xixi xjxj yiyi
37
Function G By assuming that the graph is densely connected we can make the approximation: Now G is a non-linear additive function of the beliefs:
38
Function G Instead of learningthe function can be learnt with an additive model: weighted regression stumps
39
Function G The weak learner is chosen by minimizing the loss:
40
The Boosted Random Field Algorithm For t=1..T find the best basis function h for f find the best basis function for compute local evidence compute compatibilities update the beliefs update weights xixi xjxj yiyi
41
The Boosted Random Field Algorithm For t=1..T find the best basis function h for f find the best basis function for compute local evidence compute compatibilities update the beliefs update weights xixi b1b1 b2b2 bjbj …
42
Final classifier For t=1..T update local evidences F update compatibilities G compute current beliefs Output classification:
43
Multiclass Detection U: Dictionary of ~2000 images patches V: Same number of image masks
44
Multiclass Detection U: Dictionary of ~2000 images patches V: Same number of image masks At each round t, for each class c for each dictionary entry d there is a weak learner:
45
Function f To take into account different sizes, we first downsample the image and then upsample and OR the scales: which is our function for computing the local evidence.
46
Function g The compatibily function has a similar form:
47
Function g The compatibily function has a similar form: W represent a kernel with all the messages directed to node x, y, c
48
Kernels W Example of incoming messages:
49
Function G The overall incoming messages function is given by:
50
Learning… Labeled dataset of office and street scenes, with each ~100 images In the first 5 round updated only the local evidence After the 5th iteration update also the compatibility functions At each round update only F and G of the single object class that reduces the most the multiclass cost.
51
Learning… Biggest objects are detected first because they reduce the error of all classes the fastest:
52
The End
53
Introduction Observed: Picture Dictionary: Dog P(Dog|Pic)
54
Introduction P(Head|Pic i ) P(Tail|Pic i ) P(Front Legs|Pic i ) P(Back Legs|Pic i )
55
Introduction Comp(Head, Legs) Comp(Head, Tail) Comp(F. Legs, B. Legs) Comp(Tail, Legs) Dog!
56
Introduction P(Piraña|Pic i ) Comp(Piraña, Legs)
57
Graphical Models Observation nodes y i Y y i can be a pixel or a patch
58
Graphical Models Hidden Nodes Local Evidence: X Dictionary Shord-hand
59
Graphical Models Compatibility Function: X
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.