Download presentation
Presentation is loading. Please wait.
Published byFay Harper Modified over 8 years ago
1
In: Pattern Analysis and Machine Intelligence, IEEE Transactions on, Vol. 30, Nr. 1 (2008), p. 36-51. Group of Adjacent Contour Segments for Object Detection 學生 : 戴玉書 教授 : 王聖智 老師 LEAR laboratory - INRIA Grenoble
2
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
3
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
4
Introduction Test image Training images
5
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
6
Contour segment network
7
1. Edgels extracted with Berkeley boundary detector 2. Edgel-chains partitioned into straight contour segments 3. segments connected at edgel-chains’ endpoints and junctions Contour segment network [Ferrari et al. ECCV 2006]
8
- Links between edgel-chains
9
- Connects between segments
10
Learning to Detect Natural Image Boundaries Using Local Brightness, Color, and Texture Cues [David R. Martin, Member, IEEE, Charless C. Fowlkes, and Jitendra Malik, Member, IEEE]
12
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
13
k adjacent segments (kAS) Three kinds of 2AS
14
- Descriptor of kAS
15
D(a, b) - Comparing kAS
16
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
17
kAS can be detected by a depth-first search started from every segment Detecting kAS 1 2 34 5 6
18
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
19
Using a clique-partitioning (CP) approach Let G be a complete graph whose nodes are the training kAS, and arcs are weighted by d − D(a, b) kAS codebook Each resulting clique is a cluster of similar kAS 1 2 3 D(1, 2) D(1, 3) D(2, 3)
20
- Clique-partitioning (CP) approach
21
The 35 most frequent 2AS types from the codebook The 35 most frequent 3AS from the codebook
22
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
23
|B| ・ |C| - dimensional window descriptor Object class detection -Training 1. Subdivide window into tiles 2. Compute a separate bag of PAS per tile 3. Concatenate these semi-local bags
24
- Training 1. Learn mean positive window dimensions 2. Determine number of tiles T 3. Collect positive example descriptors 4. Collect negative example descriptors: slide window over negative training images 5. Train a linear SVM
25
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
26
Object class detection -Testing 1. Slide window of aspect ratio, at multiple scales 2. SVM classify each window
27
Introduction Contour segment network k adjacent segments (kAS) -Detecting kAS -kAS CODEBOOK Object class detection -Training -Testing Result Outline
28
Result
30
Red : 2AS Blue : HoG
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.