Descriptors for CV

KNN: K- Nearest Neighbors. 39. Bag of features for image ... 41. 4. Image classification: discriminative method and much more … SVM, … 41. Bag of features for ...
8MB taille 38 téléchargements 262 vues
Descriptors for CV

Content

1.Introduction

2.Histograms

3.HOG

4.LBP

5.Haar Wavelets

6.Video based descriptor

7.How to compare descriptors

8.BoW paradigm

2014

1

T. Chateau

2

T. Chateau

1

Institut Pascal

2

Color RGB histogram

Introduc)on:     Image  descriptors   - Descriptor: a high level representation of an image or video.

- Descriptor: a vector

- Dense and sparse descriptors

- Useful for:

- geometric based matching

- image retrieval,

- object recognition and categorization,

- .... 3

T. Chateau

Institut Pascal

3

4

T. Chateau

Institut Pascal

4

Color HS+V histogram

Histogram of oriented gradients (HOG)

HS

V

5

T. Chateau

Institut Pascal

5

Combining histograms

7 Institut Pascal

7

Institut Pascal

6

Histogram of oriented gradients (HOG)

T. Chateau

6

T. Chateau

8

T. Chateau

Institut Pascal

8

LBP: local binary pattern

LBP: local binary pattern

9

T. Chateau

Institut Pascal

10

T. Chateau

9

Institut Pascal

10

Haar Wavelets

Haar Wavelets

11

T. Chateau

Institut Pascal

11

12

T. Chateau

Institut Pascal

12

Video based descriptors (HOG/HOF)

Bag of Words (bag of features models)

Origin  1:  texture  recogni)on  

13

T. Chateau

Institut Pascal

13

Bag of Words (bag of features models)

Origin  1:  texture  recogni)on  

Origin  2:  text  analysis  (Frequency  of   words  of  a  dic)onary,  Salton  &  McGill  (1983))  

15 Institut Pascal

15

Institut Pascal

14

Bag of Words (bag of features models)

T. Chateau

14

T. Chateau

16

T. Chateau

Institut Pascal

16

Bag of Words

Bag of Words

1.  extract  features  

2.  learn  visual  vocabulary  

17

T. Chateau

Institut Pascal

18

T. Chateau

17

Institut Pascal

18

Bag of Words

Bag of features for image classification

3.  represent  images  by  frequencies  of   «  visual  words  »     Given an image to classify: 1. Extract features 2. Quantize features using visual vocabulary 3. Represent images by frequencies of “visual words” 4. Estimate the class using a previously learn classifier (eg. SVM, KNN, AdaBoost,…)

19

T. Chateau

Institut Pascal

19

20

T. Chateau

Institut Pascal

20

Bag of features for image classification

Bag of features for image classification

1. Feature extraction

1. Feature extraction

Regular grid – Vogel & Schiele, 2003 – Fei-Fei & Perona, 2005

Regular grid – Vogel & Schiele, 2003 – Fei-Fei & Perona, 2005 Interest point detector – Csurka et al. 2004 – Fei-Fei & Perona, 2005 – Sivic et al. 2005

21

T. Chateau

Institut Pascal

22

T. Chateau

Institut Pascal

21

22

Bag of features for image classification

Bag of features for image classification

1. Feature extraction

1. Feature extraction

Slide credit: Josef Sivic T. Chateau

23 Institut Pascal

23

Slide credit: Josef Sivic T. Chateau

24 Institut Pascal

24

Bag of features for image classification

Bag of features for image classification

2. Learning the visual vocabulary

2. Learning the visual vocabulary

Slide credit: Josef Sivic T. Chateau

25 Institut Pascal

Slide credit: Josef Sivic T. Chateau

25

26

Bag of features for image classification

Bag of features for image classification

2. Learning the visual vocabulary

2. Learning the visual vocabulary

K-Means Clustering

Slide credit: Josef Sivic T. Chateau

27 Institut Pascal

27

26 Institut Pascal

The codebook is used for quantizing features – A vector quantizer takes a feature vector and maps it to the index of the nearest codevector in a codebook – Codebook = visual vocabulary – Codevector = visual word

28

T. Chateau

Institut Pascal

28

Bag of features for image classification

Bag of features for image classification 2. Learning the visual vocabulary exemple of visual vocabulary

2. Learning the visual vocabulary

The codebook is used for quantizing features – A vector quantizer takes a feature vector and maps it to the index of the nearest codevector in a codebook – Codebook = visual vocabulary – Codevector = visual word

Fei-Fei et al. 2005 29

T. Chateau

Institut Pascal

30

T. Chateau

29

Institut Pascal

30

Bag of features for image classification

Bag of features for image classification

2. Learning the visual vocabulary exemple of visual words

3. Image representation exemple of codewords

Sivic et al. 2005 31

T. Chateau

Institut Pascal

31

32

T. Chateau

Institut Pascal

32

Bag of features for image classification

Bag of features for image classification

4. Image classification

4. Image classification: discriminative method

codewords space

33

T. Chateau

Institut Pascal

34

T. Chateau

33

Institut Pascal

34

Bag of features for image classification

Bag of features for image classification

4. Image classification: discriminative method

4. Image classification: discriminative method Minskowski-form distance:

CHI2 distance: Codewords

Codeword space

We need a to define a distance between codewords We need a to define a distance between codewords 35

T. Chateau

Institut Pascal

35

36

T. Chateau

Institut Pascal

36

Bag of features for image classification

Bag of features for image classification

4. Image classification: discriminative method

4. Image classification: discriminative method

Kullback-Leiber divergence

Battacharyya distance

We need a to define a distance between codewords 37

T. Chateau

Institut Pascal

37

Bag of features for image classification

4. Image classification: discriminative method

4. Image classification: discriminative method

KNN: K- Nearest Neighbors

linear classifiers

39 Institut Pascal

39

Institut Pascal

38

Bag of features for image classification

T. Chateau

38

T. Chateau

40

T. Chateau

Institut Pascal

40

Bag of features for image classification

Bag of features for image classification

4. Image classification: discriminative method

Results from Pascal VOC Challenge 2010

and much more … ! SVM, …

41

T. Chateau

Institut Pascal

42

T. Chateau

41

Institut Pascal

42

Bag of features for image classification

Bag of features for image classification

AR Applications

AR Applications

43

T. Chateau

Institut Pascal

43

44

T. Chateau

Institut Pascal

44

Bag of features for image classification

Bag of features for image classification

AR Applications

AR Applications

45

T. Chateau

Institut Pascal

45

46

T. Chateau

Institut Pascal

46