1. Sharing features for multi-class object detection Antonio Torralba, Kevin Murphy and Bill Freeman MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) Sept. 1, 2004

2. The lead author: Antonio Torralba

3. Goal We want a machine to be able to identify thousands of different objects as it looks around the world.

4. Multi-class object detection: Local features no car Classifier P( car | v p ) VpVp no cow Classifier P( cow | v p ) no person Classifier P(person | v p ) … Bookshelf Desk Screen Desired detector outputs: One patch

5. Need to detect as well as to recognize Detection: localize in the image screens, keyboards and bottles In detection, the big problem is to differentiate the sparse objects against the background Recognition: name each object In recognition, the problem is to discriminate between objects.

6. There are efficient solutions for detecting a single object category and view: Viola & Jones 2001; Papageorgiou & Poggio 2000, … Lowe, 1999 detecting particular objects:. But the problem of multi-class and multi-view object detection is still largely unsolved. Leibe & Schiele, 2003; detecting objects in isolation

7. Why multi-object detection is a hard problem viewpoints Need to detect Nclasses * Nviews * Nstyles. Lots of variability within classes, and across viewpoints. Object classes Styles, lighting conditions, etc, etc, etc…

8. Existing approaches for multiclass object detection (vision community) Using a set of independent binary classifiers is the dominant strategy: Viola-Jones extension for dealing with rotations - two cascades for each view Schneiderman-Kanade multiclass object detection a) One detector for each class

9. Promising approaches… Fei-Fei, Fergus, & Perona, 2003 Look for a vocabulary of edges that reduces the number of features Krempp, Geman, & Amit, 2002

10. Characteristics of one-vs-all multiclass approaches: cost Computational cost grows linearly with Nclasses * Nviews * Nstyles … Surely, this will not scale well to 30,000 object classes.

11. Characteristics of one-vs-all multiclass approaches: representation Some of these parts can only be used for this object. What is the best representation to detect a traffic sign? Very regular object: template matching will do the job Parts derived from training a binary classifier. ~100% detection rate with 0 false alarms

12. Meaningful parts, in one-vs-all approaches Part-based object representation (looking for meaningful parts): A. Agarwal and D. Roth Ullman, Vidal-Naquet, and Sali, 2004: features of intermediate complexity are most informative for (single-object) classification. M. Weber, M. Welling and P. Perona …

13. Multi-class classifiers (machine learning community) Error correcting output codes (Dietterich & Bakiri, 1995; …) But only use classification decisions (1/-1), not real values. Reducing multi-class to binary (Allwein et al, 2000) Showed that the best code matrix is problem-dependent; don’t address how to design code matrix. Bunching algorithm (Dekel and Singer, 2002) Also learns code matrix and classifies, but more complicated than our algorithm and not applied to object detection. Multitask learning (Caruana, 1997; …) Train tasks in parallel to improve generalization, share features. But not applied to object detection, nor in a boosting framework.

14. Our approach Share features across objects, automatically selecting the best sharing pattern. Benefits of shared features: –Efficiency –Accuracy –Generalization ability

15. Algorithm goals, for object recognition. We want to find the vocabulary of parts that can be shared We want share across different objects generic knowledge about detecting objects (eg, from the background). We want to share computations across classes so that computational cost < O(Number of classes)

16. Object class 1 Object class 2 Object class 3 Object class 4 Total number of hyperplanes (features): 4 x 6 = 24. Scales linearly with number of classes Independent features

17. Total number of shared hyperplanes (features): 8 May scale sub-linearly with number of classes, and may generalize better. Shared features

18. Note: sharing is a graph, not a tree b R 3 b 3 3 Objects: {R, b, 3} This defines a vocabulary of parts shared across objects

19. At the algorithmic level Our approach is a variation on boosting that allows for sharing features in a natural way. So let’s review boosting (ada-boost demo)

20. Boosting demo

21. Joint boosting, outside of the context of images. Additive models for classification +1/-1 classification classes feature responses

22. Feature sharing in additive models 1)Simple to have sharing between additive models 2)Each term h m can be mapped to a single feature H 1 = G 1,2 + G 1 H 2 = G 1,2 + G 2

23. Flavors of boosting Different boosting algorithms use different loss functions or minimization procedures (Freund & Shapire, 1995; Friedman, Hastie, Tibshhirani, 1998). We base our approach on Gentle boosting: learns faster than others (Friedman, Hastie, Tibshhirani, 1998; Lienahart, Kuranov, & Pisarevsky, 2003).

24. Joint Boosting We use the exponential multi-class cost function At each boosting round, we add a function: classes classifier output for class c membership in class c, +1/-1

25. Newton’s method Treat h m as a perturbation, and expand loss J to second order in h m classifier with perturbation squared error reweighting

26. Joint Boosting weightsquared errorWeight squared error over training data

27.  a+b b vfvf Given a sharing pattern, the decision stump parameters are obtained analytically Feature output, v For a trial sharing pattern, set weak learner parameters to optimize overall classification h m (v,c)

28. Response histograms for background (blue) and class members (red) h m (v,c) k c=1 k c=2 k c=5 The constants k prevent sharing features just due to an asymmetry between positive and negative examples for each class. They only appear during training. Joint Boosting: select sharing pattern and weak learner to minimize cost. Algorithm details in CVPR 2004, Torralba, Murphy & Freeman

29. Approximate best sharing But this requires exploring 2 C –1 possible sharing patterns Instead we use a first-best search: S = [] 1) We fit stumps for each class independently 2) take best class - c i S = [S c i ] fit stumps for [S c i ] with c i not in S go to 2, until length(S) = Nclasses 3) select the sharing with smallest WLS error

30. Effect of pattern of feature sharing on number of features required (synthetic example)

32. 2-d synthetic example 3 classes + 1 background class

33. No feature sharing Three one-vs-all binary classifiers This is with only 8 separation lines

34. With feature sharing This is with only 8 separation lines Some lines can be shared across classes.

35. The shared features

36. Comparison of the classifiers Shared features: note better isolation of individual classes. Non-shared features.

37. Now, apply this to images. Image features (weak learners) 32x32 training image of an object 12x12 patch Location of that patch within the 32x32 object g f (x) Mean = 0 Energy = 1 w f (x) Binary mask Feature output

38. The candidate features template position

39. The candidate features Dictionary of 2000 candidate patches and position masks, randomly sampled from the training images template position

40. Database of 2500 images Annotated instances

41. Multiclass object detection 21 objects We use [20, 50] training samples per object, and about 20 times as many background examples as object examples.

42. Feature sharing at each boosting round during training

44. Example shared feature (weak classifier) At each round of running joint boosting on training set we get a feature and a sharing pattern. Response histograms for background (blue) and class members (red)

45. Non-shared feature Shared feature

46. Non-shared feature Shared feature

47. How the features were shared across objects (features sorted left-to-right from generic to specific)

48. Performance evaluation Area under ROC (shown is.9) False alarm rate Correct detection rate

49. Performance improvement over training Significant benefit to sharing features using joint boosting.

50. ROC curves for our 21 object database How will this work under training-starved or feature-starved conditions? Presumably, in the real world, we will always be starved for training data and for features.

51. 70 features, 20 training examples (left) Shared features Non-shared features

52. 15 features, 20 training examples (mid) 70 features, 20 training examples (left) Shared features Non-shared features

53. 15 features, 2 training examples (right) 15 features, 20 training examples (middle) 70 features, 20 training examples (left) Shared features Non-shared features

54. Scaling Joint Boosting shows sub-linear scaling of features with objects (for area under ROC = 0.9). Results averaged over 8 training sets, and different combinations of objects. Error bars show variability.

55. Red: shared features Blue: independent features

57. Examples of correct detections

58. What makes good features? Depends on whether we are doing single- class or multi-class detection…

59. Generic vs. specific features Parts derived from training a binary classifier. In both cases ~100% detection rate with 0 false alarms Parts derived from training a joint classifier with 20 more objects.

60. Qualitative comparison of features, for single-class and multi-class detectors

61. Multi-view object detection train for object and orientation View invariant features View specific features Sharing features is a natural approach to view-invariant object detection.

62. Multi-view object detection Sharing is not a tree. Depends also on 3D symmetries. … …

63. Multi-view object detection

64. Strong learner H response for car as function of assumed view angle

65. Visual summary…

66. Features Object units

68. Summary Argued that feature sharing will be an essential part of scaling up object detection to hundreds or thousands of objects (and viewpoints). We introduced joint boosting, a generalization to boosting that incorporates feature sharing in a natural way. Initial results (up to 30 objects) show the desired scaling behavior for # features vs # objects. The shared features are observed to generalize better, allowing learning from fewer examples, using fewer features.

69. end