47
Learning Object Detectors with Weak Supervision Kun He Committee members: Prof. Stan Sclaroff Prof. Margrit Betke Prof. Pedro Felzenszwalb

Learning Object Detectors with Weak Supervision

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Learning Object Detectors with Weak Supervision

Learning Object Detectors with Weak Supervision

Kun He

Committee members:

Prof. Stan Sclaroff

Prof. Margrit Betke

Prof. Pedro Felzenszwalb

Page 2: Learning Object Detectors with Weak Supervision

Problem: object detection

09/24/2014Learning Object Detectors with Weak Supervision

2

Source: The PASCAL Visual Object Classes Challenge 2007

Page 3: Learning Object Detectors with Weak Supervision

Supervised learning pipeline

09/24/2014Learning Object Detectors with Weak Supervision

3

• Image credit: Sudheendra Vijayanarasimhan

Page 4: Learning Object Detectors with Weak Supervision

What about annotations?

• Example: Microsoft COCO (Lin et al ECCV’14)

09/24/2014Learning Object Detectors with Weak Supervision

4

Page 5: Learning Object Detectors with Weak Supervision

What about annotations?

Image credit: Tsung-Yi Lin

09/24/2014Learning Object Detectors with Weak Supervision

5

Page 6: Learning Object Detectors with Weak Supervision

What about annotations?

09/24/2014Learning Object Detectors with Weak Supervision

6

Example taken from Microsoft COCO dataset http://mscoco.org/explore/?id=79387

Page 7: Learning Object Detectors with Weak Supervision

What about annotations?

09/24/2014Learning Object Detectors with Weak Supervision

7

Example taken from Microsoft COCO dataset http://mscoco.org/explore/?id=79387

Page 8: Learning Object Detectors with Weak Supervision

Relaxing annotation requirements

• Annotation process: laborious & error-prone

• Learn directly from the images! (weak supervision)

09/24/2014Learning Object Detectors with Weak Supervision

8

Page 9: Learning Object Detectors with Weak Supervision

Literature review outline

• Weber et al ECCV’00, Fergus et al CVPR’03, Crandall & Huttenlocher ECCV’06

Generative models

Discriminative: Multiple Instance Learning (MIL)

• Vijayanarasimhan & Grauman CVPR’08, Siva & Xiang ICCV’11, Cinbis et al CVPR’14, Song et al ICML’14 …MI-SVM

• Deselaers et al IJCV’12MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

9

Page 10: Learning Object Detectors with Weak Supervision

Literature review outline

• Weber et al ECCV’00, Fergus et al CVPR’03, Crandall & Huttenlocher ECCV’06

Generative models

Discriminative: Multiple Instance Learning (MIL)

• Vijayanarasimhan & Grauman CVPR’08, Siva & Xiang ICCV’11, Cinbis et al CVPR’14, Song et al ICML’14 …MI-SVM

• Deselaers et al IJCV’12MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

10

Page 11: Learning Object Detectors with Weak Supervision

Generative part-based models

• Detect sparse features → fit part-based model → determine (non-)existence of object

• Rob Fergus, Pietro Perona and Andrew Zisserman, CVPR’03

09/24/2014Learning Object Detectors with Weak Supervision

11

Page 12: Learning Object Detectors with Weak Supervision

Generative models (Fergus et al CVPR’03)

09/24/2014Learning Object Detectors with Weak Supervision

12

• Likelihood ratio test

• Likelihood: product of Gaussians• Features: location X, scale S, appearance A

• h : hypothesis (part-based object configuration)

Foreground model

Background model

Page 13: Learning Object Detectors with Weak Supervision

Generative models (Fergus et al CVPR’03)

• Learning: maximum likelihood via EM• E-step: expectation wrt. ℎ

• M-step: update Gaussian parameters

09/24/2014Learning Object Detectors with Weak Supervision

13

Page 14: Learning Object Detectors with Weak Supervision

Generative models (Fergus et al CVPR’03)

• Learning: maximum likelihood via EM• E-step: expectation wrt. ℎ

• M-step: update Gaussian parameters

09/24/2014Learning Object Detectors with Weak Supervision

14

𝑂(#𝑝𝑎𝑟𝑡𝑠#𝑓𝑒𝑎𝑡𝑢𝑟𝑒𝑠)Typical: 630

Page 15: Learning Object Detectors with Weak Supervision

Generative models (Fergus et al CVPR’03)

• Learned 6-part model for “face”

09/24/2014Learning Object Detectors with Weak Supervision

15

Page 16: Learning Object Detectors with Weak Supervision

Generative models (Fergus et al CVPR’03)

• Face: single-Gaussian appearance model fails

09/24/2014Learning Object Detectors with Weak Supervision

16

Page 17: Learning Object Detectors with Weak Supervision

Generative models (Fergus et al CVPR’03)

• Spotted cat: single-Gaussian shape model fails

09/24/2014Learning Object Detectors with Weak Supervision

17

Page 18: Learning Object Detectors with Weak Supervision

Generative models: critiques

• GoodProbabilistic formulation

Models multiple factors

• Bad EM is slow

Limited modeling power

09/24/2014Learning Object Detectors with Weak Supervision

18

Page 19: Learning Object Detectors with Weak Supervision

Generative models: critiques

• GoodProbabilistic formulation

Models multiple factors

• Bad EM is slow

Limited modeling power

• Discriminative models• Only model the decision boundary

• Usually perform better, eg. DPM (Felzenszwalb et al PAMI’10)

09/24/2014Learning Object Detectors with Weak Supervision

19

Page 20: Learning Object Detectors with Weak Supervision

Literature review outline

• Weber et al ECCV’00, Fergus et al CVPR’03, Crandall & Huttenlocher ECCV’06

Generative models

Discriminative: Multiple Instance Learning (MIL)

• Vijayanarasimhan & Grauman CVPR’08, Siva & Xiang ICCV’11, Cinbis et al CVPR’14, Song et al ICML’14 …MI-SVM

• Deselaers et al IJCV’12MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

20

Page 21: Learning Object Detectors with Weak Supervision

Multiple Instance Learning (MIL)

09/24/2014Learning Object Detectors with Weak Supervision

21

Image credit: Samarjit Das

Page 22: Learning Object Detectors with Weak Supervision

Multiple Instance Learning (MIL)

• Images as bags

• Candidate generation• Segmentation [Galleguillos et al ECCV’08]

• Objectness [Alexe et al PAMI’12]

• Selective Search [Uijlings et al IJCV’13]

• EdgeBoxes [Zitnick & Dollar ECCV’14]

• ……

09/24/2014Learning Object Detectors with Weak Supervision

22

Page 23: Learning Object Detectors with Weak Supervision

MIL for learning detectors

• Chicken-and-egg problem / latent variable model

Optimize(positive_instances, model_parameters)

• EM-like algorithms (MI-SVM, MI-CRF)• Impute latent variables

• Update model parameters

• Iterate

09/24/2014Learning Object Detectors with Weak Supervision

23

latent

Page 24: Learning Object Detectors with Weak Supervision

Literature review outline

• Weber et al ECCV’00, Fergus et al CVPR’03, Crandall & Huttenlocher ECCV’06

Generative models

Discriminative: Multiple Instance Learning (MIL)

• Vijayanarasimhan & Grauman CVPR’08, Siva & Xiang ICCV’11, Cinbis et al CVPR’14, Song et al ICML’14 …MI-SVM

• Deselaers et al IJCV’12MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

24

Page 25: Learning Object Detectors with Weak Supervision

SVM review

09/24/2014Learning Object Detectors with Weak Supervision

25

Page 26: Learning Object Detectors with Weak Supervision

MI-SVM

09/24/2014Learning Object Detectors with Weak Supervision

26

Page 27: Learning Object Detectors with Weak Supervision

MI-SVM

• “Witness”: identified positive instance within a positive bag

09/24/2014Learning Object Detectors with Weak Supervision

27

Page 28: Learning Object Detectors with Weak Supervision

MI-SVM algorithm (Andrews et al NIPS’02)

09/24/2014Learning Object Detectors with Weak Supervision

28

1. Initialize

2. Update witnesses for positive bags•

3. Update model• solve fully-supervised SVM

4. Repeat

• Convergence: to local optimum

Page 29: Learning Object Detectors with Weak Supervision

Progression of MI-SVM

• Source: R. Gokberk Cinbis, Jakob Verbeek and Cordelia Schmid, CVPR’14

09/24/2014Learning Object Detectors with Weak Supervision

29

Page 30: Learning Object Detectors with Weak Supervision

MI-SVM: critiques

• GoodSimple optimization problem, solvers available

• Bad Sensitive to initialization

Witness update: no strong coupling between images

09/24/2014Learning Object Detectors with Weak Supervision

30

Page 31: Learning Object Detectors with Weak Supervision

Literature review outline

• Weber et al ECCV’00, Fergus et al CVPR’03, Crandall & Huttenlocher ECCV’06

Generative models

Discriminative: Multiple Instance Learning (MIL)

• Vijayanarasimhan & Grauman CVPR’08, Siva & Xiang ICCV’11, Cinbis et al CVPR’14, Song et al ICML’14 …MI-SVM

• Deselaers et al IJCV’12MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

31

Page 32: Learning Object Detectors with Weak Supervision

Conditional Random Fields (CRF) for MIL

• Enforce similarity between witnesses

09/24/2014Learning Object Detectors with Weak Supervision

32

Page 33: Learning Object Detectors with Weak Supervision

MI-CRF (Deselaers et al IJCV’12)

• Pairwise CRF

09/24/2014Learning Object Detectors with Weak Supervision

33

“objectness”

similarity

Page 34: Learning Object Detectors with Weak Supervision

MI-CRF (Deselaers et al IJCV’12)

• Pairwise CRF

• “Objectness”

• Ω: generic “objectness”

• Π: class-specific shape score

• Υ: class-specific appearance score

• Similarity

• Λ: shape similarity

• Γ: appearance similarity

09/24/2014Learning Object Detectors with Weak Supervision

34

Page 35: Learning Object Detectors with Weak Supervision

MI-CRF: algorithm

09/24/2014Learning Object Detectors with Weak Supervision

35

Localize objects by

optimizing global energy

Page 36: Learning Object Detectors with Weak Supervision

MI-CRF: results• Example detections

• Models learned by DPM (Felzenszwalb et al PAMI’10) vs. MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

36

Page 37: Learning Object Detectors with Weak Supervision

MI-CRF: critiques

• GoodStrong coupling between images

• Bad High complexity (fully-connected CRF)

Limited #candidates per image (<=100)

09/24/2014Learning Object Detectors with Weak Supervision

37

Page 38: Learning Object Detectors with Weak Supervision

Literature review outline

• Weber et al ECCV’00, Fergus et al CVPR’03, Crandall & Huttenlocher ECCV’06

Generative models

Discriminative: Multiple Instance Learning (MIL)

• Vijayanarasimhan & Grauman CVPR’08, Siva & Xiang ICCV’11, Cinbis et al CVPR’14, Song et al ICML’14 …MI-SVM

• Deselaers et al IJCV’12MI-CRF

09/24/2014Learning Object Detectors with Weak Supervision

38

Page 39: Learning Object Detectors with Weak Supervision

Beyond MIL

• OPTIMOL: Li et al CVPR’07

• NEIL: Chen et al ICCV’13

Active learning

• Improving MI-SVM

Current research

09/24/2014Learning Object Detectors with Weak Supervision

39

Page 40: Learning Object Detectors with Weak Supervision

Active learning

• Closing the loop

09/24/2014Learning Object Detectors with Weak Supervision

40

?

Page 41: Learning Object Detectors with Weak Supervision

OPTIMOL (automatic Object Picture collecTion via Incremental MOdel Learning)

• Li-Jia Li, Gang Wang and Li Fei-Fei, CVPR’07

09/24/2014Learning Object Detectors with Weak Supervision

41

Page 42: Learning Object Detectors with Weak Supervision

NEIL (Never-Ending Image Learner)

09/24/2014Learning Object Detectors with Weak Supervision

42

• Xinlei Chen, Abhinav Shrivastava and Abhinav Gupta, ICCV’13

Page 43: Learning Object Detectors with Weak Supervision

Current research: improving MI-SVM

• MI-SVM (→ local optimum)1. Update witnesses independently

2. Update model parameters: solve SVM

• Idea: relax step 1 to

Still have convergence

Freedom to enforce desired properties

09/24/2014Learning Object Detectors with Weak Supervision

43

Page 44: Learning Object Detectors with Weak Supervision

Current research: improving MI-SVM

• Enforcing similarity between witnesses

• Step t:

• Comparison: PASCAL VOC 2007, detection mAPcat cow dog

• MI-SVM 23.83, Ours 24.12

09/24/2014Learning Object Detectors with Weak Supervision

44

MI-SVM 34.8 43.7 22.2 10.4 7.8 36.2 22.0 20.6 11.1 21.4 28.7 38.0 19.6 23.7 19.8 35.4 9.8

Ours 38.9 42.4 22.5 10.4 10.6 38.3 17.2 28.0 14.5 18.9 23.4 35.6 18.8 23.2 20.3 35.8 11.3

Page 45: Learning Object Detectors with Weak Supervision

Cats

09/24/2014Learning Object Detectors with Weak Supervision

45

Page 46: Learning Object Detectors with Weak Supervision

And dogs

09/24/2014Learning Object Detectors with Weak Supervision

46

Page 47: Learning Object Detectors with Weak Supervision

Summary

• Weakly supervised object detector learning

• Existing methods• Generative

• MI-SVM

• MI-CRF

• Future directions• Active learning (eg. OPTIMOL, NEIL)

• Current research: improving MI-SVM

• Open questions: part-based, multi-modal data, etc.

09/24/2014Learning Object Detectors with Weak Supervision

47