cvpr cvpr2013 cvpr2013-158 knowledge-graph by maker-knowledge-mining

158 cvpr-2013-Exploring Weak Stabilization for Motion Feature Extraction


Source: pdf

Author: Dennis Park, C. Lawrence Zitnick, Deva Ramanan, Piotr Dollár

Abstract: We describe novel but simple motion features for the problem of detecting objects in video sequences. Previous approaches either compute optical flow or temporal differences on video frame pairs with various assumptions about stabilization. We describe a combined approach that uses coarse-scale flow and fine-scale temporal difference features. Our approach performs weak motion stabilization by factoring out camera motion and coarse object motion while preserving nonrigid motions that serve as useful cues for recognition. We show results for pedestrian detection and human pose estimation in video sequences, achieving state-of-the-art results in both. In particular, given a fixed detection rate our method achieves a five-fold reduction in false positives over prior art on the Caltech Pedestrian benchmark. Finally, we perform extensive diagnostic experiments to reveal what aspects of our system are crucial for good performance. Proper stabilization, long time-scale features, and proper normalization are all critical.

Reference: text


Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 com Abstract We describe novel but simple motion features for the problem of detecting objects in video sequences. [sent-5, score-0.478]

2 Previous approaches either compute optical flow or temporal differences on video frame pairs with various assumptions about stabilization. [sent-6, score-0.813]

3 We describe a combined approach that uses coarse-scale flow and fine-scale temporal difference features. [sent-7, score-0.492]

4 Our approach performs weak motion stabilization by factoring out camera motion and coarse object motion while preserving nonrigid motions that serve as useful cues for recognition. [sent-8, score-1.433]

5 We show results for pedestrian detection and human pose estimation in video sequences, achieving state-of-the-art results in both. [sent-9, score-0.374]

6 In this work, we explore the motion counterpart for object detection in video. [sent-17, score-0.368]

7 We show that one can exploit simple motion features to significantly increase detection accuracy with little additional computation. [sent-18, score-0.424]

8 We classify image motion into three types using a stationary world coordinate frame and a moving object coordinate frame. [sent-20, score-0.432]

9 Camera-centric motion is the movement of the camera with respect to the world. [sent-21, score-0.397]

10 Objectcentric motion is the movement of the object centroid with respect to the world. [sent-22, score-0.328]

11 Finally, part-centric motion is the movement of object parts with respect to the object. [sent-23, score-0.356]

12 com Figure 1: Illustration of various types of video stabilization: (a) no stabilization, (b) camera motion stabilization, (c) object-centric motion stabilization, (d) camera and objectcentric motion stabilization, and (e) full stabilization of camera, object-centric, and part-centric motion. [sent-28, score-1.58]

13 We posit that for detecting articulated objects such as people the majority of useful motion information is contained in partcentric motion. [sent-29, score-0.576]

14 A simple approach is to directly compute image motion features on raw video. [sent-32, score-0.398]

15 Methods that define motion features using optical flow or spacetime gradients often take this route [29]. [sent-34, score-0.799]

16 One can partly remove camera motion by looking at differences offlow [8]. [sent-35, score-0.424]

17 A more direct approach is to simply compute motion features on a stationary camera, such as [27]. [sent-36, score-0.439]

18 Such motion features encode both object- and part-centric motion. [sent-37, score-0.367]

19 When the camera is moving, one may try to register frames using a homography or egomotion estimation [18, 19], which removes some camera-centric motion but can be challenging for dynamic scenes or those with complex 3D geometry. [sent-39, score-0.494]

20 Finally, other techniques compute optical flow in an object-centric coordinate frame [13]; Figure 1(c) shows that such an approach actually encodes both camera- and part-centric motion. [sent-40, score-0.448]

21 222888888200 In this paper, we posit (and verify by experiment) that the majority of useful motion information for detecting articulated objects such as people is contained in part-centric motion. [sent-41, score-0.484]

22 To allow the temporal features to easily extract part-centric motion information, we attempt to stabilize both camera and object-centric motion, Figure 1(d). [sent-43, score-0.709]

23 We accomplish this by using coarse-scale optical flow to align a sequence of image frames. [sent-44, score-0.339]

24 Weak stabilization using coarse-scale flow has the benefit of aligning large objects such as the background or a person’s body without removing detailed motion such as an object’s parts, Figure 1(d,e). [sent-45, score-0.984]

25 While artifacts may exists around large flow discontinuities, we demonstrate that coarse-scale flow is robust in practice. [sent-46, score-0.426]

26 We use temporal difference features to capture the partcentric motion that remains after weak stabilization. [sent-47, score-0.794]

27 While features based on fine-scale optical flow [13, 8] may be extracted from the stabilized frames, fine-scale flow is notoriously difficult to extract for small parts such as arms [4]. [sent-48, score-0.759]

28 We demonstrate that when sampled at the proper temporal intervals, simple temporal difference features are an effective alternative capable of achieving state-of-the-art results. [sent-49, score-0.627]

29 We perform a thorough evaluation of motion features for object detection in video. [sent-50, score-0.473]

30 We focus on detecting pedestrians in moving cameras [12] as well as pose estimation from static cameras [1]. [sent-51, score-0.34]

31 We demonstrate significant improvements from integrating our motion features into three distinct approaches: rigid SVM detectors defined on HOG features [7], articulated part models defined on HOG features [16, 3 1], and boosted detectors defined on channel features [11]. [sent-52, score-0.77]

32 Related Work Optical-flow-based features: A popular strategy for video-based recognition is to extend static image features into the temporal domain through use of optical flow. [sent-57, score-0.597]

33 Examples include spatially blurred flow fields [13] or histograms of optical flow vectors [8, 28]. [sent-58, score-0.554]

34 For stationary cameras, temporal difference features can be computed on background models, yielding background-subtraction masks [24]. [sent-69, score-0.406]

35 Our approach can be seen as a combination of optical-flow and temporal differencing as we compute differences on spacetime windows that are weakly-stabilized with coarse optical flow. [sent-70, score-0.661]

36 Action classification: Many of the above motion features have been explored in the context of action classification [10, 21]. [sent-71, score-0.412]

37 In particular, [29] performs a thorough evaluation of motion descriptors, discovering that histograms of flow perform well. [sent-72, score-0.577]

38 Tracking: An alternate use of temporal information to improve detection reliability is to explicitly track objects. [sent-75, score-0.287]

39 Most trackers tend to define motion models on static image features, although exceptions do exist [15]. [sent-77, score-0.468]

40 We then describe our approach to weakly-stabilizing video frames and our resulting motion features. [sent-83, score-0.486]

41 Static features: In addition to the motion features introduced below, we use one of two sets of static features densely computed on the current frame. [sent-89, score-0.64]

42 Our first set of static features are the channel features described in [11]. [sent-90, score-0.405]

43 (b) Using fine-scale LK flows, the overall body is stabilized onto the last frame at the cost of distortion in body parts (most visible at the heads and legs of the top row). [sent-95, score-0.385]

44 Our second type of static features is the commonly used Histogram of Oriented Gradients (HOG) descriptor [7]. [sent-98, score-0.303]

45 Stabilizing videos Our goal is to compute motion features based on partcentric motion, such as the movement of a person’s limbs. [sent-102, score-0.537]

46 This requires weakly stabilizing image frames to remove both camera and object-centric motion while preserving the part-centric motion. [sent-103, score-0.642]

47 We accomplish this by using coarsescale optical flow to align a sequence of frames. [sent-104, score-0.339]

48 We estimate optical flow using the approach of LucasKanade [22] but applied in a somewhat non-standard manner. [sent-105, score-0.307]

49 Lucas-Kanade proposed a differential approach to flow bilized and weakly stabilized frames spaced one frame apart (m = 1) and 8 frames apart (m = 8). [sent-106, score-0.914]

50 With larger frame spans (m = 8) temporal differences appear. [sent-108, score-0.414]

51 However, × weak stabilization is needed to remove non-informative differences resulting from camera and object motion. [sent-109, score-0.628]

52 In practice, we find Wt,t−1 stabilizes the majority of motion due to camera and objectcentric motion, as shown in Figure 2. [sent-121, score-0.459]

53 Computing the coarse flows is fast (no need to compute flow at finest scale) and fairly robust (due to the large σ). [sent-122, score-0.361]

54 When stabilizing across multiple frames, we compute the global motion Wt,t−n by progressively warping and summing pairwise flow fields. [sent-123, score-0.626]

55 We found this to work better in practice than computing the potentially large flow directly between frames It and It−n. [sent-124, score-0.357]

56 Motion features Given (weakly) stabilized image frames, we propose the use of simple temporal differencing or temporal gradient features. [sent-127, score-0.82]

57 The temporal gradient is defined as the difference between two frames, Dσ = It − It−1,t, (1) where σ is the scale of the computed flow. [sent-129, score-0.35]

58 Because σ is tuned to be roughly the size of an object, we expect the temporal gradient to contain useful cues about nonrigid object motion that are helpful for detection, as in Figure 3. [sent-130, score-0.551]

59 We denote temporal gradient on unstabilized frames as DUS: DUS = It − It−1 (2) Using multiple frames: We previously defined the difference features over pairs of frames. [sent-131, score-0.549]

60 In many instances, the amount of motion observed between subsequent frames may be quite small, especially with slow moving objects. [sent-132, score-0.425]

61 First, we consider the simple approach of computing multiple frame differences between the current frame and k = n/m other frames spaced apart temporally by m frames from t − m to t − n. [sent-136, score-0.725]

62 Another approach is to compute the set of differences between neighboring frames within a multiframe set, mean frame Mt and the neighboring frames, Rectified features: Previously, we defined our temporal difference features using the signed temporal gradient. [sent-143, score-1.151]

63 Sev- ×× × eral other possibilities also exist for encoding the temporal differences, such as using the absolute value of the temporal gradient or using rectified gradients. [sent-144, score-0.57]

64 Rectified gradients compute two features for each pixel’s temporal gradient dt corresponding to max(0, dt) and max(0, −dt). [sent-145, score-0.465]

65 We begin by exploring the feature parameter space on the task of pedestrian detection using a boosting classifier [11]. [sent-166, score-0.375]

66 the frame skip m, (b) other forms of stabilization, (c) frame skip m vs. [sent-178, score-0.534]

67 frame span n, (d) various types of reference frames for computing D(m, n), (e) different types of rectification for utilizing the color channels, and (f) boosting vs. [sent-179, score-0.508]

68 The best results, D106 (8, 4), are achieved using σ = 16, m = 4, n = 8, the current frame as reference, and the signed temporal differences of the luminance channel. [sent-181, score-0.545]

69 We measure accuracy using the standard log-average miss rate for the detections [12], which is computed by averaging the miss rate at nine false positives per image (FPPI) rates evenly spaced between 10−2 to A detection is labeled as correct if the area of overlap is greater than 50%. [sent-184, score-0.491]

70 [11], as reported in [12], is a 56% log-average miss rate using only static features and trained on the INRIA dataset [7]. [sent-187, score-0.404]

71 We perform our sweeps using boosting and the 10 static channel features described in Section 3. [sent-195, score-0.508]

72 Lastly, we combine the optimal temporal features found for boosting with the static HOG features for use by linear SVMs. [sent-198, score-0.703]

73 frame skip: We first explore the space of two parameters; the scale of LK flows, σ, and the skip between two frames used to compute the temporal difference, m, see Fig. [sent-203, score-0.733]

74 When the pair of frames are temporally nearby, stabilization plays a smaller role, since objects are relatively well aligned even without stabilization. [sent-208, score-0.621]

75 As we increase the skip m between the pair of frames, stabilization becomes critical. [sent-209, score-0.586]

76 Ideally, the optical flow scale should roughly cover an object, and so would be defined relative to the size of the candidate window being evaluated. [sent-211, score-0.338]

77 Multiframe: Given a fixed scale σ = 16, we now examine the question of the optimal multiframe span n, skip m, and reference frame. [sent-218, score-0.4]

78 We find that a large span n = 8 and small skip value m = 1 performs best, although a larger skip m = 4 also does well, see Fig. [sent-220, score-0.362]

79 T Ihis yields the final multiframe motion feature of D0 (n = 8, m = 4). [sent-225, score-0.403]

80 4(e), using three temporal differences across the LUV color channels. [sent-227, score-0.304]

81 The “Max” scheme uses the maximum temporal difference across the 3 channels, while the “Lum” scheme just uses the luminance (L) channel. [sent-228, score-0.335]

82 The normalization has minimal effect on the performance of the boosting classifier, presumably because boosting classifiers can train more flexible decision boundaries that perform implicit normalization. [sent-235, score-0.297]

83 5 we compare with previous work including ‘MultiFtr+Motion’ [28] (which uses motion features) and ‘MultiresC’ [23] (which uses static features trained on the same data as [12]). [sent-239, score-0.585]

84 6 shows several examples of detections using our approach compared to using static features alone. [sent-243, score-0.326]

85 Several false detections are removed around the car’s boundary as temporal features remove the ambiguities. [sent-244, score-0.419]

86 Temporal features can also help discover missed detections, such as the pedestrian riding a bicycle in the second row. [sent-245, score-0.29]

87 Our new temporal features lead to a significant improvement across all FPPI rates. [sent-251, score-0.316]

88 The annotated frames are evenly split into training and testing, and used to evaluate the ability of our motion features to perform human pose estimation in video sequences. [sent-256, score-0.653]

89 Baseline articulated part model: We describe our baseline articulated part model [3 1], and show how to extend it to incorporate our motion features. [sent-257, score-0.413]

90 Motion features: For our experiments, we simply augment the appearance descriptor to include both HOG and 222888888755 models; one trained only with static features (left), and the other trained with both static and our motion features (right). [sent-266, score-0.919]

91 Note that our motion features help detect instances that are considered hard due to abnormal pose (biking) or occlusion, and significantly reduce false positives. [sent-267, score-0.469]

92 (7) The above formulation allows us to easily incorporate our motion features into the existing pipeline at both test-time and train-time. [sent-270, score-0.367]

93 We show estimates from the pose model of [3 1] trained using our motion features. [sent-276, score-0.364]

94 Multiple people often interact and occlude each other, making pose estimation and motion extraction difficult. [sent-283, score-0.377]

95 Conclusion We described a family of temporal features utilizing weakly stabilized video frames. [sent-285, score-0.581]

96 Weak stabilization enables our detectors to easily extract part-centric informa- tion by removing most camera- and object-centric motion. [sent-286, score-0.429]

97 We experimentally show that simple temporal differences extracted across large time-spans are capable of producing 222888888866 Table1:LUAFoupOwegHvamrtuealdnrtmsgi a67n35012Ha. [sent-287, score-0.304]

98 25t0M 3m% otdienlwhur motion features produces consistently better part localizations. [sent-289, score-0.367]

99 Large displacement optical flow: descriptor matching in variational motion estimation. [sent-320, score-0.405]

100 Human detection using oriented histograms of flow and appearance. [sent-340, score-0.304]


similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('stabilization', 0.429), ('motion', 0.281), ('temporal', 0.23), ('flow', 0.213), ('pedestrian', 0.204), ('static', 0.187), ('skip', 0.157), ('caltech', 0.148), ('frames', 0.144), ('stabilized', 0.125), ('mindseye', 0.122), ('multiframe', 0.122), ('boosting', 0.114), ('frame', 0.11), ('objectcentric', 0.109), ('differencing', 0.109), ('stabilizing', 0.101), ('optical', 0.094), ('partcentric', 0.092), ('spacetime', 0.087), ('features', 0.086), ('flows', 0.081), ('hog', 0.08), ('doll', 0.076), ('sweeps', 0.075), ('signed', 0.075), ('differences', 0.074), ('miss', 0.07), ('rectified', 0.07), ('normalization', 0.069), ('camera', 0.069), ('articulated', 0.066), ('tpami', 0.062), ('video', 0.061), ('body', 0.061), ('spaced', 0.059), ('fppi', 0.059), ('channels', 0.058), ('lk', 0.057), ('detection', 0.057), ('weak', 0.056), ('luminance', 0.056), ('hof', 0.053), ('detections', 0.053), ('pose', 0.052), ('pedestrians', 0.051), ('rectification', 0.05), ('detecting', 0.05), ('false', 0.05), ('thorough', 0.049), ('difference', 0.049), ('temporally', 0.048), ('span', 0.048), ('irvine', 0.047), ('dus', 0.047), ('movement', 0.047), ('weakly', 0.047), ('channel', 0.046), ('deva', 0.045), ('action', 0.045), ('uc', 0.044), ('people', 0.044), ('stabilize', 0.043), ('posit', 0.043), ('walk', 0.043), ('positives', 0.043), ('person', 0.042), ('reference', 0.042), ('schindler', 0.042), ('stationary', 0.041), ('dt', 0.04), ('gradient', 0.04), ('shrinking', 0.038), ('gradients', 0.038), ('dalal', 0.036), ('coarse', 0.036), ('refers', 0.036), ('apart', 0.036), ('toe', 0.036), ('mt', 0.035), ('pooling', 0.034), ('wojek', 0.034), ('histograms', 0.034), ('ramanan', 0.033), ('boosted', 0.033), ('proper', 0.032), ('family', 0.032), ('accomplish', 0.032), ('scale', 0.031), ('trained', 0.031), ('tth', 0.031), ('compute', 0.031), ('jones', 0.03), ('explore', 0.03), ('pages', 0.03), ('descriptor', 0.03), ('rate', 0.03), ('evenly', 0.029), ('hours', 0.029), ('parts', 0.028)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.99999964 158 cvpr-2013-Exploring Weak Stabilization for Motion Feature Extraction

Author: Dennis Park, C. Lawrence Zitnick, Deva Ramanan, Piotr Dollár

Abstract: We describe novel but simple motion features for the problem of detecting objects in video sequences. Previous approaches either compute optical flow or temporal differences on video frame pairs with various assumptions about stabilization. We describe a combined approach that uses coarse-scale flow and fine-scale temporal difference features. Our approach performs weak motion stabilization by factoring out camera motion and coarse object motion while preserving nonrigid motions that serve as useful cues for recognition. We show results for pedestrian detection and human pose estimation in video sequences, achieving state-of-the-art results in both. In particular, given a fixed detection rate our method achieves a five-fold reduction in false positives over prior art on the Caltech Pedestrian benchmark. Finally, we perform extensive diagnostic experiments to reveal what aspects of our system are crucial for good performance. Proper stabilization, long time-scale features, and proper normalization are all critical.

2 0.32894567 333 cvpr-2013-Plane-Based Content Preserving Warps for Video Stabilization

Author: Zihan Zhou, Hailin Jin, Yi Ma

Abstract: Recently, a new image deformation technique called content-preserving warping (CPW) has been successfully employed to produce the state-of-the-art video stabilization results in many challenging cases. The key insight of CPW is that the true image deformation due to viewpoint change can be well approximated by a carefully constructed warp using a set of sparsely constructed 3D points only. However, since CPW solely relies on the tracked feature points to guide the warping, it works poorly in large textureless regions, such as ground and building interiors. To overcome this limitation, in this paper we present a hybrid approach for novel view synthesis, observing that the textureless regions often correspond to large planar surfaces in the scene. Particularly, given a jittery video, we first segment each frame into piecewise planar regions as well as regions labeled as non-planar using Markov random fields. Then, a new warp is computed by estimating a single homography for regions belong to the same plane, while in- heriting results from CPW in the non-planar regions. We demonstrate how the segmentation information can be efficiently obtained and seamlessly integrated into the stabilization framework. Experimental results on a variety of real video sequences verify the effectiveness of our method.

3 0.26782751 244 cvpr-2013-Large Displacement Optical Flow from Nearest Neighbor Fields

Author: Zhuoyuan Chen, Hailin Jin, Zhe Lin, Scott Cohen, Ying Wu

Abstract: We present an optical flow algorithm for large displacement motions. Most existing optical flow methods use the standard coarse-to-fine framework to deal with large displacement motions which has intrinsic limitations. Instead, we formulate the motion estimation problem as a motion segmentation problem. We use approximate nearest neighbor fields to compute an initial motion field and use a robust algorithm to compute a set of similarity transformations as the motion candidates for segmentation. To account for deviations from similarity transformations, we add local deformations in the segmentation process. We also observe that small objects can be better recovered using translations as the motion candidates. We fuse the motion results obtained under similarity transformations and under translations together before a final refinement. Experimental validation shows that our method can successfully handle large displacement motions. Although we particularly focus on large displacement motions in this work, we make no sac- rifice in terms of overall performance. In particular, our method ranks at the top of the Middlebury benchmark.

4 0.25738162 334 cvpr-2013-Pose from Flow and Flow from Pose

Author: Katerina Fragkiadaki, Han Hu, Jianbo Shi

Abstract: Human pose detectors, although successful in localising faces and torsos of people, often fail with lower arms. Motion estimation is often inaccurate under fast movements of body parts. We build a segmentation-detection algorithm that mediates the information between body parts recognition, and multi-frame motion grouping to improve both pose detection and tracking. Motion of body parts, though not accurate, is often sufficient to segment them from their backgrounds. Such segmentations are crucialfor extracting hard to detect body parts out of their interior body clutter. By matching these segments to exemplars we obtain pose labeled body segments. The pose labeled segments and corresponding articulated joints are used to improve the motion flow fields by proposing kinematically constrained affine displacements on body parts. The pose-based articulated motion model is shown to handle large limb rotations and displacements. Our algorithm can detect people under rare poses, frequently missed by pose detectors, showing the benefits of jointly reasoning about pose, segmentation and motion in videos.

5 0.22957294 59 cvpr-2013-Better Exploiting Motion for Better Action Recognition

Author: Mihir Jain, Hervé Jégou, Patrick Bouthemy

Abstract: Several recent works on action recognition have attested the importance of explicitly integrating motion characteristics in the video description. This paper establishes that adequately decomposing visual motion into dominant and residual motions, both in the extraction of the space-time trajectories and for the computation of descriptors, significantly improves action recognition algorithms. Then, we design a new motion descriptor, the DCS descriptor, based on differential motion scalar quantities, divergence, curl and shear features. It captures additional information on the local motion patterns enhancing results. Finally, applying the recent VLAD coding technique proposed in image retrieval provides a substantial improvement for action recognition. Our three contributions are complementary and lead to outperform all reported results by a significant margin on three challenging datasets, namely Hollywood 2, HMDB51 and Olympic Sports. 1. Introduction and related work Human actions often convey the essential meaningful content in videos. Yet, recognizing human actions in un- constrained videos is a challenging problem in Computer Vision which receives a sustained attention due to the potential applications. In particular, there is a large interest in designing video-surveillance systems, providing some automatic annotation of video archives as well as improving human-computer interaction. The solutions proposed to address this problem inherit, to a large extent, from the techniques first designed for the goal of image search and classification. The successful local features developed to describe image patches [15, 23] have been translated in the 2D+t domain as spatio-temporal local descriptors [13, 30] and now include motion clues [29]. These descriptors are often extracted from spatial-temporal interest points [12, 3 1]. More recent techniques assume some underlying temporal motion model involving trajectories [2, 6, 7, 17, 18, 25, 29, 32]. Most of these approaches produce large set of local descriptors which are in turn aggregated to produce a single vector representing the video, in order to enable the use of powerful discriminative classifiers such as support vector machines (SVMs). This is usually done with the bag- Figure 1. Optical flow field vectors (green vectors with red end points) before and after dominant motion compensation. Most of the flow vectors due to camera motion are suppressed after compensation. One of the contributions of this paper is to show that compensating for the dominant motion is beneficial for most of the existing descriptors used for action recognition. of-words technique [24], which quantizes the local features using a k-means codebook. Thanks to the successful combination of this encoding technique with the aforementioned local descriptors, the state of the art in action recognition is able to go beyond the toy problems ofclassifying simple human actions in controlled environment and considers the detection of actions in real movies or video clips [11, 16]. Despite these progresses, the existing descriptors suffer from an uncompleted handling of motion in the video sequence. Motion is arguably the most reliable source of information for action recognition, as often related to the actions of interest. However, it inevitably involves the background or camera motion when dealing with uncontrolled and re- alistic situations. Although some attempts have been made to compensate camera motion in several ways [10, 21, 26, 29, 32], how to separate action motion from that caused by the camera, and how to reflect it in the video description remains an open issue. The motion compensation mechanism employed in [10] is tailor-made to the Motion Interchange Pattern encoding technique. The Motion Boundary Histogram (MBH) [29] is a recent appealing approach to 222555555533 suppress the constant motion by considering the flow gradient. It is robust to some extent to the presence of camera motion, yet it does not explicitly handle the camera motion. Another approach [26] uses a sophisticated and robust (RANSAC) estimation of camera motion. It first segments the color image into regions corresponding to planar parts in the scene and estimates the (three) dominant homographies to update the motion associated with local features. A rather different view is adopted in [32] where the motion decomposition is performed at the trajectory level. All these works support the potential of motion compensation. As the first contribution of this paper, we address the problem in a way that departs from these works by considering the compensation of the dominant motion in both the tracking stages and encoding stages involved in the computation of action recognition descriptors. We rely on the pioneering works on motion compensation such as the technique proposed in [20], that considers 2D polynomial affine motion models for estimating the dominant image motion. We consider this particular model for its robustness and its low computational cost. It was already used in [21] to separate the dominant motion (assumed to be due to the camera motion) and the residual motion (corresponding to the independent scene motions) for dynamic event recognition in videos. However, the statistical modeling of both motion components was global (over the entire image) and only the normal flow was computed for the latter. Figure 1 shows the vectors of optical flow before and after applying the proposed motion compensation. Our method successfully suppresses most of the background motion and reinforces the focus towards the action of interest. We exploit this compensated motion both for descriptor computation and for extracting trajectories. However, we also show that the camera motion should not be thrown as it contains complementary information that is worth using to recognize certain action categories. Then, we introduce the Divergence-Curl-Shear (DCS) descriptor, which encodes scalar first-order motion features, namely the motion divergence, curl and shear. It captures physical properties of the flow pattern that are not involved in the best existing descriptors for action recognition, except in the work of [1] which exploits divergence and vorticity among a set of eleven kinematic features computed from the optical flow. Our DCS descriptor provides a good performance recognition performance on its own. Most importantly, it conveys some information which is not captured by existing descriptors and further improves the recognition performance when combined with the other descriptors. As a last contribution, we bring an encoding technique known as VLAD (vector oflocal aggregated descriptors) [8] to the field of action recognition. This technique is shown to be better than the bag-of-words representation for combining all the local video descriptors we have considered. The organization of the paper is as follows. Section 2 introduces the motion properties that we will consider through this paper. Section 3 presents the datasets and classification scheme used in our different evaluations. Section 4 details how we revisit several popular descriptors of the literature by the means of dominant motion compensation. Our DCS descriptor based on kinematic properties is introduced in Section 5 and improved by the VLAD encoding technique, which is introduced and bench-marked in Section 6 for several video descriptors. Section 7 provides a comparison showing the large improvement achieved over the state of the art. Finally, Section 8 concludes the paper. 2. Motion Separation and Kinematic Features In this section, we describe the motion clues we incorporate in our action recognition framework. We separate the dominant motion and the residual motion. In most cases, this will account to distinguishing the impact of camera movement and independent actions. Note that we do not aim at recovering the 3D camera motion: The 2D parametric motion model describes the global (or dominant) motion between successive frames. We first explain how we estimate the dominant motion and employ it to separate the dominant flow from the optical flow. Then, we will introduce kinematic features, namely divergence, curl and shear for a more comprehensive description of the visual motion. 2.1. Affine motion for compensating camera motion Among polynomial motion models, we consider the 2D affine motion model. Simplest motion models such as the 4parameter model formed by the combination of 2D translation, 2D rotation and scaling, or more complex ones such as the 8-parameter quadratic model (equivalent to a homography), could be selected as well. The affine model is a good trade-off between accuracy and efficiency which is of primary importance when processing a huge video database. It does have limitations since strictly speaking it implies a single plane assumption for the static background. However, this is not that penalizing (especially for outdoor scenes) if differences in depth remain moderated with respect to the distance to the camera. The affine flow vector at point p = (x, y) and at time t, is defined as waff(pt) =?cc12((t ) ?+?aa31((t ) aa42((t ) ? ?xytt?. (1) = + + = + uaff(pt) c1(t) a1(t)xt a2(t)yt and vaff(pt) c2(t) a3 (t)xt + a4(t)yt are horizontal and vertical components of waff(pt) respectively. Let us denote the optical flow vector at point p at time t as w(pt) = (u(pt) , v(pt)). We introduce the flow vector ω(pt) obtained by removing the affine flow vector from the optical flow vector ω(pt) = w(pt) − waff(pt) . (2) 222555555644 The dominant motion (estimated as waff(pt)) is usually due to the camera motion. In this case, Equation 2 amounts to canceling (or compensating) the camera motion. Note that this is not always true. For example in case of close-up on a moving actor, the dominant motion will be the affine estimation of the apparent actor motion. The interpretation of the motion compensation output will not be that straightforward in this case, however the resulting ω-field will still exhibit different patterns for the foreground action part and the background part. In the remainder, we will refer to the “compensated” flow as ω-flow. Figure 1 displays the computed optical flow and the ωflow. We compute the affine flow with the publicly available Motion2D software1 [20] which implements a realtime robust multiresolution incremental estimation framework. The affine motion model has correctly accounted for the motion induced by the camera movement which corresponds to the dominant motion in the image pair. Indeed, we observe that the compensated flow vectors in the background are close to null and the compensated flow in the foreground, i.e., corresponding to the actors, is conversely inflated. The experiments presented along this paper will show that effective separation of dominant motion from the residual motions is beneficial for action recognition. As explained in Section 4, we will compute local motion descriptors, such as HOF, on both the optical flow and the compensated flow (ω-flow), which allows us to explicitly and directly characterize the scene motion. 2.2. Local kinematic features By kinematic features, we mean local first-order differential scalar quantities computed on the flow field. We consider the divergence, the curl (or vorticity) and the hyperbolic terms. They inform on the physical pattern of the flow so that they convey useful information on actions in videos. They can be computed from the first-order derivatives of the flow at every point p at every frame t as ⎨⎪ ⎪ ⎪ ⎧hcdyuipvr1l2(p t) = −∂ u ∂ (yxp(xtp) +−∂ v ∂ v(px ypxt ) The diverg⎪⎩ence is related to axial motion, expansion scaling effects, the curl to rotation in the image plane. hyperbolic terms express the shear of the visual flow responding to more complex configuration. We take account the shear quantity only: shear(pt) = ?hyp12(pt) + hyp22(pt). (3) and The corinto (4) 1http://www.irisa.fr/vista/Motion2D/ In Section 5, we propose the DCS descriptor that is based on the kinematic features (divergence, curl and shear) of the visual motion discussed in this subsection. It is computed on either the optical or the compensated flow, ω-flow. 3. Datasets and evaluation This section first introduces the datasets used for the evaluation. Then, we briefly present the bag-of-feature model and the classification scheme used to encode the descriptors which will be introduced in Section 4. Hollywood2. The Hollywood2 dataset [16] contains 1,707 video clips from 69 movies representing 12 action classes. It is divided into train set and test set of 823 and 884 samples respectively. Following the standard evaluation protocol of this benchmark, we use average precision (AP) for each class and the mean of APs (mAP) for evaluation. HMDB51. The HMDB51 dataset [11] is a large dataset containing 6,766 video clips extracted from various sources, ranging from movies to YouTube. It consists of 51 action classes, each having at least 101 samples. We follow the evaluation protocol of [11] and use three train/test splits, each with 70 training and 30 testing samples per class. The average classification accuracy is computed over all classes. Out of the two released sets, we use the original set as it is more challenging and used by most of the works reporting results in action recognition. Olympic Sports. The third dataset we use is Olympic Sports [19], which again is obtained from YouTube. This dataset contains 783 samples with 16 sports action classes. We use the provided2 train/test split, there are 17 to 56 training samples and 4 to 11test samples per class. Mean AP is used for the evaluation, which is the standard choice. Bag of features and classification setup. We first adopt the standard BOF [24] approach to encode all kinds of descriptors. It produces a vector that serves as the video representation. The codebook is constructed for each type of descriptor separately by the k-means algorithm. Following a common practice in the literature [27, 29, 30], the codebook size is set to k=4,000 elements. Note that Section 6 will consider encoding technique for descriptors. For the classification, we use a non-linear SVM with χ2kernel. When combining different descriptors, we simply add the kernel matrices, as done in [27]: K(xi,xj) = exp?−?cγ1cD(xic,xjc)?, 2http://vision.stanford.edu/Datasets/OlympicSports/ 222555555755 (5) where D(xic, xjc) is χ2 distance between video xic and xjc with respect to c-th channel, corresponding to c-th descriptor. The quantity γc is the mean value of χ2 distances between the training samples for the c-th channel. The multiclass classification problem that we consider is addressed by applying a one-against-rest approach. 4. Compensated descriptors This section describes how the compensation ofthe dominant motion is exploited to improve the quality of descriptors encoding the motion and the appearance around spatio-temporal positions, hence the term “compensated descriptors”. First, we briefly review the local descriptors [5, 13, 16, 29, 30] used here along with dense trajectories [29]. Second, we analyze the impact of motion flow compensation when used in two different stages of the descriptor computation, namely in the tracking and the description part. 4.1. Dense trajectories and local descriptors Employing dense trajectories to compute local descriptors is one of the state-of-the-art approaches for action recognition. It has been shown [29] that when local descriptors are computed over dense trajectories the performance improves considerably compared to when computed over spatio temporal features [30]. Dense Trajectories [29]: The trajectories are obtained by densely tracking sampled points using optical flow fields. First, feature points are sampled from a dense grid, with step size of 5 pixels and over 8 scales. Each feature point pt = (xt, yt) at frame t is then tracked to the next frame by median filtering in a dense optical flow field F = (ut, vt) as follows: pt+1 = (xt+1 , yt+1) = (xt, yt) + (M ∗ F) | (x ¯t,y ¯t) , (6) where M is the kernel of median filtering and ( x¯ t, y¯ t) is the rounded position of (xt, yt). The tracking is limited to L (=15) frames to avoid any drifting effect. Excessively short trajectories and trajectories exhibiting sudden large displacements are removed as they induce some artifacts. Trajectories must be understood here as tracks in the spacetime volume of the video. Local descriptors: The descriptors are computed within a space-time volume centered around each trajectory. Four types of descriptors are computed to encode the shape of the trajectory, local motion pattern and appearance, namely Trajectory [29], HOF (histograms of optical flow) [13], MBH [4] and HOG (histograms of oriented gradients) [3]. All these descriptors depend on the flow field used for the tracking and as input of the descriptor computation: 1. The Trajectory descriptor encodes the shape of the trajectory represented by the normalized relative coor- × dinates of the successive points forming the trajectory. It directly depends on the dense flow used for tracking points. 2. HOF is computed using the orientations and magnitudes of the flow field. 3. MBH is designed to capture the gradient of horizontal and vertical components of the flow. The motion boundaries encode the relative pixel motion and therefore suppress camera motion, but only to some extent. 4. HOG encodes the appearance by using the intensity gradient orientations and magnitudes. It is formally not a motion descriptor. Yet the position where the descriptor is computed depends on the trajectory shape. As in [29], volume around a feature point is divided into a 2 2 3 space-time grid. The orientations are quantized ian 2to × ×8 b2i ×ns 3fo srp HacOe-Gti amned g g9r ibdi.ns T fhoer o oHriOenFt (awtioitnhs one a qdudainttiiozneadl zero bin). The horizontal and vertical components of MBH are separately quantized into 8 bins each. 4.2. Impact of motion compensation The optical flow is simply referred to as flow in the following, while the compensated flow (see subsection 2. 1) is denoted by ω-flow. Both of them are considered in the tracking and descriptor computation stages. The trajectories obtained by tracking with the ω-flow are called ω-trajectories. Figure 2 comparatively illustrates the ωtrajectories and the trajectories obtained using the flow. The input video shows a man moving away from the car. In this video excerpt, the camera is following the man walking to the right, thus inducing a global motion to the left in the video. When using the flow, the computed trajectories reflect the combination of these two motion components (camera and scene motion) as depicted by Subfigure 2(b), which hampers the characterization of the current action. In contrast, the ω-trajectories plotted in Subfigure 2(c) are more active on the actor moving on the foreground, while those localized in the background are now parallel to the time axis enhancing static parts of the scene. The ω-trajectories are therefore more relevant for action recognition, since they are more regularly and more exclusively following the actor’s motion. Impact on Trajectory and HOG descriptors. Table 1reports the impact of ω-trajectories on Trajectory and HOG descriptors, which are both significantly improved by 3%4% of mAP on the two datasets. When improved by ωflow, these descriptors will be respectively referred to as ω-Trajdesc and ω-HOG in the rest of the paper. Although the better performance of ω-Trajdesc versus the original Trajectory descriptor was expected, the one 222555555866 2. Trajectories obtained from optical and compensated flows. The green tail is the trajectory the current frame. The trajectories are sub-sampled for the sake of clarity. The frames are extracted Figure over every 15 frames with red dot indicating 5 frames in this example. DescriptorHollywood2HMDB51 BaseTrliaωnje- Tc(rtoarejrdpyreos[c2d9u]ced)54 7 1. 7 4% %2382.–89% BaseliHnωOe- (GHreOp [2rG9od]uced)4 451 . 658%%%2296.– 13%% Table 1. ω-Trajdesc and ω-HOG: Impact of compensating flow on Trajectory descriptor and HOG descriptors. achieved by ω-HOG might be surprising. Our interpretation is that HOG captures more context with the modified trajectories. More precisely, the original HOG descriptor is computed from a 2D+t sub-volume aligned with the corresponding trajectory and hence represents the appearance along the trajectory shape. When using ω-flow, we do not align the video sequence. As a result, the ω-HOG descriptor is no more computed around the very same tracked physical point in the space-time volume but around points lying in a patch of the initial feature point, whose size depends on the affine flow magnitude. ω-HOG can be viewed as a “patchbased” computation capturing more information about the appearance of the background or of the moving foreground. As for ω-trajectories, they are closer to the real trajectories of the moving actors as they usually cancel the camera movement, and so, more easier to train and recognize. Impact on HOF. The ω-flow impacts computation used as an input to HOF computation itself. Therefore, HOF can both types of trajectories (ω-trajectories both the trajectory and the descriptor be computed along or those extracted MethodHollywood2HMDB51 Table(ω2rHf.-alocO IwomkF)inpHgacOtFobf[2u9ωhsb]i:f-nlo ωgwot-hwωHOflFown5H 02 34O. 58291F% %descripto3 r706s38.:–1076% m%APfor Hollywood2 and average accuracy for HMDB5 1. The ω-HOF is used in subsequent evaluations. from flow) and can encode both kinds of flows (ω-flow or flow). For the sake of completeness, we evaluate all the variants as well as the combination of both flows in the descriptor computation stage. The results are presented in Table 2 and demonstrate the significant improvement obtained by computing the HOF descriptor with the ω-flow instead of the optical flow. Note that the type of trajectories which is used, either “Tracking flow” or “Tracking ω-flow”, has a limited impact in this case. From now on, we only consider the “Tracking ω-flow” case where HOF is computed along ω-trajectories. Interestingly, combining the HOF computed from the flow and the ω-flow further improves the results. This suggests that the two flow fields are complementary and the affine flow that was subtracted from ω-flow brings in additional information. For the sake of brevity, the combination of the two kinds of HOF, i.e., computed from the flow and the ω-flow using ω-trajectories, is referred to as the ω-HOF 222555555977 MethodHollywood2HMDB51 Tab(lerT3a.cIkmM inpBgMacHgωtBf-loH w [u2)s9in]gω f-lo wo MBH5 d42 e.052s7c% riptos:m34A90P.–3769f% orHllywood2 and average accuracy for HMDB5 1. DTerHasMjcBrOeblitpHGeFor4.ySumTωraw- frcilykto hw ionfgtheduωpCs-fcaolrtωmeiwNp-df/tl+Aωoutrw-finlogwthdesωcr- isTpc-fHtrMloaiOjrBpdswtGeHFosrc descriptor in the rest of this paper. Compared to the HOF baseline, the ω-HOF descriptor achieves a gain of +3.1% of mAP on Hollywood 2 and of +7.8% on HMDB51. Impact on MBH. Since MBH is computed from gradient of flow and cancel the constant motion, there is practically no benefit in using the ω-flow to compute the MBH descriptors, as shown in Table 3. However, by tracking ω-flow, the performance improves by around 1.3% for HMDB5 1 dataset and drops by around 1.5% for Hollywood2. This relative performance depends on the encoding technique. We will come back on this descriptor when considering another encoding scheme for local descriptors in Section 6. 4.3. Summary of compensated descriptors Table 4 summarizes the refined versions of the descriptors obtained by exploiting the ω-flow, and both ω-flow and the optical flow in the case of HOF. The revisited descriptors considerably improve the results compared to the orig- inal ones, with the noticeable exception of ω-MBH which gives mixed performance with a bag-of-features encoding scheme. But we already mention as this point that this incongruous behavior of ω-MBH is stabilized with the VLAD encoding scheme considered in Section 6. Another advantage of tracking the compensated flow is that fewer trajectories are produced. For instance, the total number of trajectories decreases by about 9. 16% and 22.81% on the Hollywood2 and HMDB51 datasets, respectively. Note that exploiting both the flow and the ω-flow do not induce much computational overhead, as the latter is obtained from the flow and the affine flow which is computed in real-time and already used to get the ω-trajectories. The only additional computational cost that we introduce by using the descriptors summarized in Table 4 is the computation of a second HOF descriptor, but this stage is relatively efficient and not the bottleneck of the extraction procedure. 5. Divergence-Curl-Shear descriptor This section introduces a new descriptor encoding the kinematic properties of motion discussed in Section 2.2. It is denoted by DCS in the rest of this paper. Combining kinematic features. The spatial derivatives are computed for the horizontal and vertical components of the flow field, which are used in turn to compute the divergence, curl and shear scalar values, see Equation 3. We consider all possible pairs of kinematic features, namely (div, curl), (div, shear) and (curl, shear). At each × ×× pixel, we compute the orientation and magnitude of the 2-D vector corresponding to each of these pairs. The orientation is quantized into histograms and the magnitude is used for weighting, similar to SIFT. Our motivation for encoding pairs is that the joint distribution of kinematic features conveys more information than exploiting them independently. Implementation details. The descriptor computation and parameters are similar to HOG and other popular descriptors such as MBH, HOF. We obtain 8-bin histograms for each of the three feature pairs or components of DCS. The range of possible angles is 2π for the (div,curl) pair and π for the other pairs, because the shear is always positive. The DCS descriptor is computed for a space-time volume aligned with a trajectory, as done with the four descriptors mentioned in the previous section. In order to capture the spatio-temporal structure of kinematic features, the volume (32 32 pixels and L = 15 frames) is subdivided into a spatio-temporal grid nofd s Lize = nx 5× f ny m×e nt, sw situhb nx =de ny =to 2a and nt = 3. These parameters ×hnave× × bneen fixed for the sake of consistency with the other descriptors. For each pair of kinematic features, each cell in the grid is represented by a histogram. The resulting local descriptors have a dimensionality equal to 288 = nx ny nt 8 3. At the video level, these descriptors are nenc×od end i×nto 8 a single vector representation using either BOF or the VLAD encoding scheme introduced in the next section. 6. VLAD in actions VLAD [8] is a descriptor encoding technique that aggregates the descriptors based on a locality criterion in the feature space. To our knowledge, this technique has never been considered for action recognition. Below, we briefly introduce this approach and give the performance achieved for all the descriptors introduced along the previous sections. VLAD in brief. Similar to BOF, VLAD relies on a codebook C = {c1, c2 , ...ck} of k centroids learned by k-means. bTohoek representation is ob}t oaifn ked c by summing, efodr b yea kch-m mveiasunasl. word ci, the differences x − ci of the vectors x assigned to ci, thereby producing a sv exct −or c representation oflength d×k, 222555556088 DMeBscHriptorV5 LH.A1o%Dlywo5Bo4d.O2 %F4V3L.3HA%MD B35B91.O7%F Taωbl-eDHM5rOCBa.FSjGPdHe+rsωfco-mMHaBOnFeofV54L2936A.51D% with5431ω208-.5T96% rajde3s42c97158,.ω3% -HOG342,58019ω.6-% HOF descriptors and their combination. where d is the dimension ofthe local descriptors. We use the codebook size, k = 256. Despite this large dimensionality, VLAD is efficient because it is effectively compared with a linear kernel. VLAD is post-processed using a componentwise power normalization, which dramatically improves its performance [8]. While cross validating the parameter α involved in this power normalization, we consistently observe, for all the descriptors, a value between 0.15 and 0.3. Therefore, this parameter is set to α = 0.2 in all our experiments. For classification, we use a linear SVM and oneagainst-rest approach everywhere, unless stated otherwise. Impact on existing descriptors. We employ VLAD because it is less sensitive to quantization parameters and appears to provide better performance with descriptors having a large dimensionality. These properties are interesting in our case, because the quantization parameters involved in the DCS and MBH descriptors have been used unchanged in Section 4 for the sake of direct comparison. They might be suboptimal when using the ω-flow instead of the optical flow on which they have initially been optimized [29]. Results for MBH and ω-MBH in Table 5 supports this argument. When using VLAD instead of BOF, the scores are stable in both the cases and there is no mixed inference as that observed in Table 3. VLAD also has significant positive influence on accuracy of ω-DCS descriptor. We also observe that ω-DCS is complementary to ω-MBH and adds to the performance. Still DCS is probably not best utilized in the current setting of parameters. In case of ω-Trajdesc and ω-HOG, the scores are better with BOF on both the datasets. ω-HOF with VLAD improves on HMDB5 1, but remains equivalent for Hollywood2. Although BOF leads to better scores for the descriptors considered individually, their combination with VLAD outperforms the BOF. 7. Comparison with the state of the art This section reports our results with all descriptors combined and compares our method with the state of the art. TrajectorCy+omHbOiGna+tHioOnF+ MDBCHSHol5 l98y.w76%o%od2H4M489.D02%B%51 All ω-descriptors all five compensated descriptors using combined62.5%52.1% Table 6. Combination of VLAD representation. WU*JVliaOnughreM tHaeolth. [yo2w9d87o] 256 0985. 37% SKa*duOeJinhrau tnegdMteHatlMa.h [ol1Dd.0B[91]25 24 609.8172% Table 7. Comparison with the state of the art on Hollywood2 and HMDB5 1 datasets. *Vig et al. [28] gets 61.9% by using external eye movements data. *Jiang et al. [9] used one-vs-one multi class SVM while our and other methods use one-vs-rest SVMs. With one-against-one multi class SVM we obtain 45. 1% for HMDB51. Descriptor combination. Table 6 reports the results obtained when the descriptors are combined. Since we use VLAD, our baseline is updated that is combination of Trajectory, HOG, HOF and MBH with VLAD representation. When DCS is added to the baseline there is an improvement of 0.9% and 1.2%. With combination of all five compensated descriptors we obtain 62.5% and 52.1% on the two datasets. This is a large improvement even over the updated baseline, which shows that the proposed motion compensation and the way we exploit it are significantly important for action recognition. The comparison with the state of the art is shown in Table 7. Our method outperforms all the previously reported results in the literature. In particular, on the HMDB51 dataset, the improvement over the best reported results to date is more than 11% in average accuracy. Jiang el al. [9] used a one-against-one multi-class SVM, which might have resulted in inferior scores. With a similar multi-class SVM approach, our method obtains 45. 1%, which remains significantly better than their result. All others results were reported with one-against-rest approach. On Olympic Sports dataset we obtain mAP of 83.2% with ‘All ω-descriptors combined’ and the improvement is mostly because of VLAD and ω-flow. The best reported mAPs on this dataset are Liu et al. [14] (74.4%) and Jiang et al. [9] (80.6%), which we exceed convincingly. Gaidon et al. [6] reports the best average accuracy of 82.7%. 8. Conclusions This paper first demonstrates the interest of canceling the dominant motion (predominantly camera motion) to make the visual motion truly related to actions, for both the trajectory extraction and descriptor computation stages. It pro222555556199 duces significantly better versions (called compensated descriptors) of several state-of-the-art local descriptors for action recognition. The simplicity, efficiency and effectiveness of this motion compensation approach make it applicable to any action recognition framework based on motion descriptors and trajectories. The second contribution is the new DCS descriptor derived from the first-order scalar motion quantities specifying the local motion patterns. It captures additional information which is proved complementary to the other descriptors. Finally, we show that VLAD encoding technique instead of bag-of-words boosts several action descriptors, and overall exhibits a significantly better performance when combining different types of descriptors. Our contributions are all complementary and significantly outperform the state of the art when combined, as demon- strated by our extensive experiments on the Hollywood 2, HMDB51 and Olympic Sports datasets. Acknowledgments This work was supported by the Quaero project, funded by Oseo, French agency for innovation. We acknowledge Heng Wang’s help for reproducing some of their results. References [1] S. Ali and M. Shah. Human action recognition in videos using kinematic features and multiple instance learning. IEEE T-PAMI, 32(2):288–303, Feb. 2010. [2] T. Brox and J. Malik. Object segmentation by long term analysis of point trajectories. In ECCV, Sep. 2010. [3] N. Dalal and B. Triggs. Histograms of oriented gradients for human detection. In CVPR, Jun. 2005. [4] N. Dalal, B. Triggs, and C. Schmid. Human detection using oriented histograms of flow and appearance. In ECCV, May 2006. [5] P. Dollar, V. Rabaud, G. Cottrell, and S. Belongie. Behavior recognition via sparse spatio-temporal features. In VS-PETS, Oct. 2005. [6] A. Gaidon, Z. Harchaoui, and C. Schmid. Recognizing activities with cluster-trees of tracklets. In BMVC, Sep. 2012. [7] A. Hervieu, P. Bouthemy, and J.-P. Le Cadre. A statistical video content recognition method using invariant features on object trajectories. IEEE T-CSVT, 18(1 1): 1533–1543, 2008. [8] H. J ´egou, F. Perronnin, M. Douze, J. S ´anchez, P. P ´erez, and C. Schmid. Aggregating local descriptors into compact [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] codes. IEEE T-PAMI, 34(9):1704–1716, 2012. Y.-G. Jiang, Q. Dai, X. Xue, W. Liu, and C.-W. Ngo. Trajectory-based modeling of human actions with motion reference points. In ECCV, Oct. 2012. O. Kliper-Gross, Y. Gurovich, T. Hassner, and L. Wolf. Motion interchange patterns for action recognition in unconstrained videos. In ECCV, Oct. 2012. H. Kuehne, H. Jhuang, E. Garrote, T. Poggio, and T. Serre. Hmdb: A large video database for human motion recognition. In ICCV, Nov. 2011. I. Laptev and T. Lindeberg. Space-time interest points. In ICCV, Oct. 2003. I. Laptev, M. Marzalek, C. Schmid, and B. Rozenfeld. Learning realistic human actions from movies. In CVPR, Jun. 2008. J. Liu, B. Kuipers, and S. Savarese. Recognizing human actions by attributes. In CVPR, Jun. 2011. D. Lowe. Distinctive image features from scale-invariant keypoints. IJCV, 60(2):91–1 10, Nov. 2004. M. Marzalek, I. Laptev, and C. Schmid. Actions in context. In CVPR, Jun. 2009. P. Matikainen, M. Hebert, and R. Sukthankar. Trajectons: Action recognition through the motion analysis of tracked features. In Workshop on Video-Oriented Object and Event Classification, ICCV, Sep. 2009. R. Messing, C. J. Pal, and H. A. Kautz. Activity recognition using the velocity histories of tracked keypoints. In ICCV, Sep. 2009. J. C. Niebles, C.-W. Chen, and F.-F. Li. Modeling temporal structure of decomposable motion segments for activity classification. In ECCV, Sep. 2010. [20] J.-M. Odobez and P. Bouthemy. Robust multiresolution estimation of parametric motion models. Jal of Vis. Comm. and Image Representation, 6(4):348–365, Dec. 1995. [21] G. Piriou, P. Bouthemy, and J.-F. Yao. Recognition of dynamic video contents with global probabilistic models of visual motion. IEEE T-IP, 15(1 1):3417–3430, 2006. [22] S. Sadanand and J. J. Corso. Action bank: A high-level representation of activity in video. In CVPR, Jun. 2012. [23] C. Schmid and R. Mohr. Local grayvalue invariants for image retrieval. IEEE T-PAMI, 19(5):530–534, May 1997. [24] J. Sivic and A. Zisserman. Video Google: A text retrieval approach to object matching in videos. In ICCV, pages 1470– 1477, Oct. 2003. [25] J. Sun, X. Wu, S. Yan, L. F. Cheong, T.-S. Chua, and J. Li. Hierarchical spatio-temporal context modeling for action recognition. In CVPR, Jun. 2009. [26] H. Uemura, S. Ishikawa, and K. Mikolajczyk. Feature tracking and motion compensation for action recognition. In BMVC, Sep. 2008. [27] M. M. Ullah, S. N. Parizi, and I. Laptev. Improving bag-offeatures action recognition with non-local cues. In BMVC, Sep. 2010. [28] E. Vig, M. Dorr, and D. Cox. Saliency-based space-variant descriptor sampling for action recognition. In ECCV, Oct. 2012. [29] H. Wang, A. Kl¨ aser, C. Schmid, and C.-L. Liu. Action recognition by dense trajectories. In CVPR, Jun. 2011. [30] H. Wang, M. M. Ullah, A. Kl¨ aser, I. Laptev, and C. Schmid. Evaluation of local spatio-temporal features for action recognition. In BMVC, Sep. 2009. [3 1] G. Willems, T. Tuytelaars, and L. J. V. Gool. An efficient dense and scale-invariant spatio-temporal interest point detector. In ECCV, Oct. 2008. [32] S. Wu, O. Oreifej, and M. Shah. Action recognition in videos acquired by a moving camera using motion decomposition of lagrangian particle trajectories. In ICCV, Nov. 2011. 222555666200

6 0.22294456 398 cvpr-2013-Single-Pedestrian Detection Aided by Multi-pedestrian Detection

7 0.21770689 363 cvpr-2013-Robust Multi-resolution Pedestrian Detection in Traffic Scenes

8 0.20843789 124 cvpr-2013-Determining Motion Directly from Normal Flows Upon the Use of a Spherical Eye Platform

9 0.1873181 332 cvpr-2013-Pixel-Level Hand Detection in Ego-centric Videos

10 0.18651286 187 cvpr-2013-Geometric Context from Videos

11 0.16996393 383 cvpr-2013-Seeking the Strongest Rigid Detector

12 0.16812333 10 cvpr-2013-A Fully-Connected Layered Model of Foreground and Background Flow

13 0.16620338 386 cvpr-2013-Self-Paced Learning for Long-Term Tracking

14 0.16049603 345 cvpr-2013-Real-Time Model-Based Rigid Object Pose Estimation and Tracking Combining Dense and Sparse Visual Cues

15 0.1591661 328 cvpr-2013-Pedestrian Detection with Unsupervised Multi-stage Feature Learning

16 0.1546981 40 cvpr-2013-An Approach to Pose-Based Action Recognition

17 0.15345436 137 cvpr-2013-Dynamic Scene Classification: Learning Motion Descriptors with Slow Features Analysis

18 0.14632316 459 cvpr-2013-Watching Unlabeled Video Helps Learn New Human Actions from Very Few Labeled Snapshots

19 0.14588456 300 cvpr-2013-Multi-target Tracking by Lagrangian Relaxation to Min-cost Network Flow

20 0.14254361 203 cvpr-2013-Hierarchical Video Representation with Trajectory Binary Partition Tree


similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, 0.309), (1, 0.044), (2, 0.017), (3, -0.186), (4, -0.147), (5, -0.014), (6, 0.13), (7, -0.077), (8, -0.005), (9, 0.083), (10, 0.03), (11, 0.078), (12, 0.224), (13, -0.171), (14, 0.246), (15, 0.072), (16, -0.03), (17, -0.057), (18, -0.108), (19, -0.0), (20, -0.063), (21, -0.093), (22, -0.093), (23, -0.011), (24, -0.073), (25, -0.015), (26, -0.019), (27, 0.093), (28, 0.016), (29, 0.055), (30, -0.025), (31, 0.023), (32, 0.042), (33, 0.066), (34, -0.039), (35, 0.016), (36, -0.015), (37, 0.051), (38, -0.11), (39, -0.011), (40, -0.002), (41, -0.027), (42, -0.09), (43, -0.005), (44, 0.087), (45, -0.024), (46, -0.037), (47, 0.002), (48, -0.021), (49, -0.018)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.96586901 158 cvpr-2013-Exploring Weak Stabilization for Motion Feature Extraction

Author: Dennis Park, C. Lawrence Zitnick, Deva Ramanan, Piotr Dollár

Abstract: We describe novel but simple motion features for the problem of detecting objects in video sequences. Previous approaches either compute optical flow or temporal differences on video frame pairs with various assumptions about stabilization. We describe a combined approach that uses coarse-scale flow and fine-scale temporal difference features. Our approach performs weak motion stabilization by factoring out camera motion and coarse object motion while preserving nonrigid motions that serve as useful cues for recognition. We show results for pedestrian detection and human pose estimation in video sequences, achieving state-of-the-art results in both. In particular, given a fixed detection rate our method achieves a five-fold reduction in false positives over prior art on the Caltech Pedestrian benchmark. Finally, we perform extensive diagnostic experiments to reveal what aspects of our system are crucial for good performance. Proper stabilization, long time-scale features, and proper normalization are all critical.

2 0.75979644 244 cvpr-2013-Large Displacement Optical Flow from Nearest Neighbor Fields

Author: Zhuoyuan Chen, Hailin Jin, Zhe Lin, Scott Cohen, Ying Wu

Abstract: We present an optical flow algorithm for large displacement motions. Most existing optical flow methods use the standard coarse-to-fine framework to deal with large displacement motions which has intrinsic limitations. Instead, we formulate the motion estimation problem as a motion segmentation problem. We use approximate nearest neighbor fields to compute an initial motion field and use a robust algorithm to compute a set of similarity transformations as the motion candidates for segmentation. To account for deviations from similarity transformations, we add local deformations in the segmentation process. We also observe that small objects can be better recovered using translations as the motion candidates. We fuse the motion results obtained under similarity transformations and under translations together before a final refinement. Experimental validation shows that our method can successfully handle large displacement motions. Although we particularly focus on large displacement motions in this work, we make no sac- rifice in terms of overall performance. In particular, our method ranks at the top of the Middlebury benchmark.

3 0.74253047 137 cvpr-2013-Dynamic Scene Classification: Learning Motion Descriptors with Slow Features Analysis

Author: Christian Thériault, Nicolas Thome, Matthieu Cord

Abstract: In this paper, we address the challenging problem of categorizing video sequences composed of dynamic natural scenes. Contrarily to previous methods that rely on handcrafted descriptors, we propose here to represent videos using unsupervised learning of motion features. Our method encompasses three main contributions: 1) Based on the Slow Feature Analysis principle, we introduce a learned local motion descriptor which represents the principal and more stable motion components of training videos. 2) We integrate our local motion feature into a global coding/pooling architecture in order to provide an effective signature for each video sequence. 3) We report state of the art classification performances on two challenging natural scenes data sets. In particular, an outstanding improvement of 11 % in classification score is reached on a data set introduced in 2012.

4 0.74017668 59 cvpr-2013-Better Exploiting Motion for Better Action Recognition

Author: Mihir Jain, Hervé Jégou, Patrick Bouthemy

Abstract: Several recent works on action recognition have attested the importance of explicitly integrating motion characteristics in the video description. This paper establishes that adequately decomposing visual motion into dominant and residual motions, both in the extraction of the space-time trajectories and for the computation of descriptors, significantly improves action recognition algorithms. Then, we design a new motion descriptor, the DCS descriptor, based on differential motion scalar quantities, divergence, curl and shear features. It captures additional information on the local motion patterns enhancing results. Finally, applying the recent VLAD coding technique proposed in image retrieval provides a substantial improvement for action recognition. Our three contributions are complementary and lead to outperform all reported results by a significant margin on three challenging datasets, namely Hollywood 2, HMDB51 and Olympic Sports. 1. Introduction and related work Human actions often convey the essential meaningful content in videos. Yet, recognizing human actions in un- constrained videos is a challenging problem in Computer Vision which receives a sustained attention due to the potential applications. In particular, there is a large interest in designing video-surveillance systems, providing some automatic annotation of video archives as well as improving human-computer interaction. The solutions proposed to address this problem inherit, to a large extent, from the techniques first designed for the goal of image search and classification. The successful local features developed to describe image patches [15, 23] have been translated in the 2D+t domain as spatio-temporal local descriptors [13, 30] and now include motion clues [29]. These descriptors are often extracted from spatial-temporal interest points [12, 3 1]. More recent techniques assume some underlying temporal motion model involving trajectories [2, 6, 7, 17, 18, 25, 29, 32]. Most of these approaches produce large set of local descriptors which are in turn aggregated to produce a single vector representing the video, in order to enable the use of powerful discriminative classifiers such as support vector machines (SVMs). This is usually done with the bag- Figure 1. Optical flow field vectors (green vectors with red end points) before and after dominant motion compensation. Most of the flow vectors due to camera motion are suppressed after compensation. One of the contributions of this paper is to show that compensating for the dominant motion is beneficial for most of the existing descriptors used for action recognition. of-words technique [24], which quantizes the local features using a k-means codebook. Thanks to the successful combination of this encoding technique with the aforementioned local descriptors, the state of the art in action recognition is able to go beyond the toy problems ofclassifying simple human actions in controlled environment and considers the detection of actions in real movies or video clips [11, 16]. Despite these progresses, the existing descriptors suffer from an uncompleted handling of motion in the video sequence. Motion is arguably the most reliable source of information for action recognition, as often related to the actions of interest. However, it inevitably involves the background or camera motion when dealing with uncontrolled and re- alistic situations. Although some attempts have been made to compensate camera motion in several ways [10, 21, 26, 29, 32], how to separate action motion from that caused by the camera, and how to reflect it in the video description remains an open issue. The motion compensation mechanism employed in [10] is tailor-made to the Motion Interchange Pattern encoding technique. The Motion Boundary Histogram (MBH) [29] is a recent appealing approach to 222555555533 suppress the constant motion by considering the flow gradient. It is robust to some extent to the presence of camera motion, yet it does not explicitly handle the camera motion. Another approach [26] uses a sophisticated and robust (RANSAC) estimation of camera motion. It first segments the color image into regions corresponding to planar parts in the scene and estimates the (three) dominant homographies to update the motion associated with local features. A rather different view is adopted in [32] where the motion decomposition is performed at the trajectory level. All these works support the potential of motion compensation. As the first contribution of this paper, we address the problem in a way that departs from these works by considering the compensation of the dominant motion in both the tracking stages and encoding stages involved in the computation of action recognition descriptors. We rely on the pioneering works on motion compensation such as the technique proposed in [20], that considers 2D polynomial affine motion models for estimating the dominant image motion. We consider this particular model for its robustness and its low computational cost. It was already used in [21] to separate the dominant motion (assumed to be due to the camera motion) and the residual motion (corresponding to the independent scene motions) for dynamic event recognition in videos. However, the statistical modeling of both motion components was global (over the entire image) and only the normal flow was computed for the latter. Figure 1 shows the vectors of optical flow before and after applying the proposed motion compensation. Our method successfully suppresses most of the background motion and reinforces the focus towards the action of interest. We exploit this compensated motion both for descriptor computation and for extracting trajectories. However, we also show that the camera motion should not be thrown as it contains complementary information that is worth using to recognize certain action categories. Then, we introduce the Divergence-Curl-Shear (DCS) descriptor, which encodes scalar first-order motion features, namely the motion divergence, curl and shear. It captures physical properties of the flow pattern that are not involved in the best existing descriptors for action recognition, except in the work of [1] which exploits divergence and vorticity among a set of eleven kinematic features computed from the optical flow. Our DCS descriptor provides a good performance recognition performance on its own. Most importantly, it conveys some information which is not captured by existing descriptors and further improves the recognition performance when combined with the other descriptors. As a last contribution, we bring an encoding technique known as VLAD (vector oflocal aggregated descriptors) [8] to the field of action recognition. This technique is shown to be better than the bag-of-words representation for combining all the local video descriptors we have considered. The organization of the paper is as follows. Section 2 introduces the motion properties that we will consider through this paper. Section 3 presents the datasets and classification scheme used in our different evaluations. Section 4 details how we revisit several popular descriptors of the literature by the means of dominant motion compensation. Our DCS descriptor based on kinematic properties is introduced in Section 5 and improved by the VLAD encoding technique, which is introduced and bench-marked in Section 6 for several video descriptors. Section 7 provides a comparison showing the large improvement achieved over the state of the art. Finally, Section 8 concludes the paper. 2. Motion Separation and Kinematic Features In this section, we describe the motion clues we incorporate in our action recognition framework. We separate the dominant motion and the residual motion. In most cases, this will account to distinguishing the impact of camera movement and independent actions. Note that we do not aim at recovering the 3D camera motion: The 2D parametric motion model describes the global (or dominant) motion between successive frames. We first explain how we estimate the dominant motion and employ it to separate the dominant flow from the optical flow. Then, we will introduce kinematic features, namely divergence, curl and shear for a more comprehensive description of the visual motion. 2.1. Affine motion for compensating camera motion Among polynomial motion models, we consider the 2D affine motion model. Simplest motion models such as the 4parameter model formed by the combination of 2D translation, 2D rotation and scaling, or more complex ones such as the 8-parameter quadratic model (equivalent to a homography), could be selected as well. The affine model is a good trade-off between accuracy and efficiency which is of primary importance when processing a huge video database. It does have limitations since strictly speaking it implies a single plane assumption for the static background. However, this is not that penalizing (especially for outdoor scenes) if differences in depth remain moderated with respect to the distance to the camera. The affine flow vector at point p = (x, y) and at time t, is defined as waff(pt) =?cc12((t ) ?+?aa31((t ) aa42((t ) ? ?xytt?. (1) = + + = + uaff(pt) c1(t) a1(t)xt a2(t)yt and vaff(pt) c2(t) a3 (t)xt + a4(t)yt are horizontal and vertical components of waff(pt) respectively. Let us denote the optical flow vector at point p at time t as w(pt) = (u(pt) , v(pt)). We introduce the flow vector ω(pt) obtained by removing the affine flow vector from the optical flow vector ω(pt) = w(pt) − waff(pt) . (2) 222555555644 The dominant motion (estimated as waff(pt)) is usually due to the camera motion. In this case, Equation 2 amounts to canceling (or compensating) the camera motion. Note that this is not always true. For example in case of close-up on a moving actor, the dominant motion will be the affine estimation of the apparent actor motion. The interpretation of the motion compensation output will not be that straightforward in this case, however the resulting ω-field will still exhibit different patterns for the foreground action part and the background part. In the remainder, we will refer to the “compensated” flow as ω-flow. Figure 1 displays the computed optical flow and the ωflow. We compute the affine flow with the publicly available Motion2D software1 [20] which implements a realtime robust multiresolution incremental estimation framework. The affine motion model has correctly accounted for the motion induced by the camera movement which corresponds to the dominant motion in the image pair. Indeed, we observe that the compensated flow vectors in the background are close to null and the compensated flow in the foreground, i.e., corresponding to the actors, is conversely inflated. The experiments presented along this paper will show that effective separation of dominant motion from the residual motions is beneficial for action recognition. As explained in Section 4, we will compute local motion descriptors, such as HOF, on both the optical flow and the compensated flow (ω-flow), which allows us to explicitly and directly characterize the scene motion. 2.2. Local kinematic features By kinematic features, we mean local first-order differential scalar quantities computed on the flow field. We consider the divergence, the curl (or vorticity) and the hyperbolic terms. They inform on the physical pattern of the flow so that they convey useful information on actions in videos. They can be computed from the first-order derivatives of the flow at every point p at every frame t as ⎨⎪ ⎪ ⎪ ⎧hcdyuipvr1l2(p t) = −∂ u ∂ (yxp(xtp) +−∂ v ∂ v(px ypxt ) The diverg⎪⎩ence is related to axial motion, expansion scaling effects, the curl to rotation in the image plane. hyperbolic terms express the shear of the visual flow responding to more complex configuration. We take account the shear quantity only: shear(pt) = ?hyp12(pt) + hyp22(pt). (3) and The corinto (4) 1http://www.irisa.fr/vista/Motion2D/ In Section 5, we propose the DCS descriptor that is based on the kinematic features (divergence, curl and shear) of the visual motion discussed in this subsection. It is computed on either the optical or the compensated flow, ω-flow. 3. Datasets and evaluation This section first introduces the datasets used for the evaluation. Then, we briefly present the bag-of-feature model and the classification scheme used to encode the descriptors which will be introduced in Section 4. Hollywood2. The Hollywood2 dataset [16] contains 1,707 video clips from 69 movies representing 12 action classes. It is divided into train set and test set of 823 and 884 samples respectively. Following the standard evaluation protocol of this benchmark, we use average precision (AP) for each class and the mean of APs (mAP) for evaluation. HMDB51. The HMDB51 dataset [11] is a large dataset containing 6,766 video clips extracted from various sources, ranging from movies to YouTube. It consists of 51 action classes, each having at least 101 samples. We follow the evaluation protocol of [11] and use three train/test splits, each with 70 training and 30 testing samples per class. The average classification accuracy is computed over all classes. Out of the two released sets, we use the original set as it is more challenging and used by most of the works reporting results in action recognition. Olympic Sports. The third dataset we use is Olympic Sports [19], which again is obtained from YouTube. This dataset contains 783 samples with 16 sports action classes. We use the provided2 train/test split, there are 17 to 56 training samples and 4 to 11test samples per class. Mean AP is used for the evaluation, which is the standard choice. Bag of features and classification setup. We first adopt the standard BOF [24] approach to encode all kinds of descriptors. It produces a vector that serves as the video representation. The codebook is constructed for each type of descriptor separately by the k-means algorithm. Following a common practice in the literature [27, 29, 30], the codebook size is set to k=4,000 elements. Note that Section 6 will consider encoding technique for descriptors. For the classification, we use a non-linear SVM with χ2kernel. When combining different descriptors, we simply add the kernel matrices, as done in [27]: K(xi,xj) = exp?−?cγ1cD(xic,xjc)?, 2http://vision.stanford.edu/Datasets/OlympicSports/ 222555555755 (5) where D(xic, xjc) is χ2 distance between video xic and xjc with respect to c-th channel, corresponding to c-th descriptor. The quantity γc is the mean value of χ2 distances between the training samples for the c-th channel. The multiclass classification problem that we consider is addressed by applying a one-against-rest approach. 4. Compensated descriptors This section describes how the compensation ofthe dominant motion is exploited to improve the quality of descriptors encoding the motion and the appearance around spatio-temporal positions, hence the term “compensated descriptors”. First, we briefly review the local descriptors [5, 13, 16, 29, 30] used here along with dense trajectories [29]. Second, we analyze the impact of motion flow compensation when used in two different stages of the descriptor computation, namely in the tracking and the description part. 4.1. Dense trajectories and local descriptors Employing dense trajectories to compute local descriptors is one of the state-of-the-art approaches for action recognition. It has been shown [29] that when local descriptors are computed over dense trajectories the performance improves considerably compared to when computed over spatio temporal features [30]. Dense Trajectories [29]: The trajectories are obtained by densely tracking sampled points using optical flow fields. First, feature points are sampled from a dense grid, with step size of 5 pixels and over 8 scales. Each feature point pt = (xt, yt) at frame t is then tracked to the next frame by median filtering in a dense optical flow field F = (ut, vt) as follows: pt+1 = (xt+1 , yt+1) = (xt, yt) + (M ∗ F) | (x ¯t,y ¯t) , (6) where M is the kernel of median filtering and ( x¯ t, y¯ t) is the rounded position of (xt, yt). The tracking is limited to L (=15) frames to avoid any drifting effect. Excessively short trajectories and trajectories exhibiting sudden large displacements are removed as they induce some artifacts. Trajectories must be understood here as tracks in the spacetime volume of the video. Local descriptors: The descriptors are computed within a space-time volume centered around each trajectory. Four types of descriptors are computed to encode the shape of the trajectory, local motion pattern and appearance, namely Trajectory [29], HOF (histograms of optical flow) [13], MBH [4] and HOG (histograms of oriented gradients) [3]. All these descriptors depend on the flow field used for the tracking and as input of the descriptor computation: 1. The Trajectory descriptor encodes the shape of the trajectory represented by the normalized relative coor- × dinates of the successive points forming the trajectory. It directly depends on the dense flow used for tracking points. 2. HOF is computed using the orientations and magnitudes of the flow field. 3. MBH is designed to capture the gradient of horizontal and vertical components of the flow. The motion boundaries encode the relative pixel motion and therefore suppress camera motion, but only to some extent. 4. HOG encodes the appearance by using the intensity gradient orientations and magnitudes. It is formally not a motion descriptor. Yet the position where the descriptor is computed depends on the trajectory shape. As in [29], volume around a feature point is divided into a 2 2 3 space-time grid. The orientations are quantized ian 2to × ×8 b2i ×ns 3fo srp HacOe-Gti amned g g9r ibdi.ns T fhoer o oHriOenFt (awtioitnhs one a qdudainttiiozneadl zero bin). The horizontal and vertical components of MBH are separately quantized into 8 bins each. 4.2. Impact of motion compensation The optical flow is simply referred to as flow in the following, while the compensated flow (see subsection 2. 1) is denoted by ω-flow. Both of them are considered in the tracking and descriptor computation stages. The trajectories obtained by tracking with the ω-flow are called ω-trajectories. Figure 2 comparatively illustrates the ωtrajectories and the trajectories obtained using the flow. The input video shows a man moving away from the car. In this video excerpt, the camera is following the man walking to the right, thus inducing a global motion to the left in the video. When using the flow, the computed trajectories reflect the combination of these two motion components (camera and scene motion) as depicted by Subfigure 2(b), which hampers the characterization of the current action. In contrast, the ω-trajectories plotted in Subfigure 2(c) are more active on the actor moving on the foreground, while those localized in the background are now parallel to the time axis enhancing static parts of the scene. The ω-trajectories are therefore more relevant for action recognition, since they are more regularly and more exclusively following the actor’s motion. Impact on Trajectory and HOG descriptors. Table 1reports the impact of ω-trajectories on Trajectory and HOG descriptors, which are both significantly improved by 3%4% of mAP on the two datasets. When improved by ωflow, these descriptors will be respectively referred to as ω-Trajdesc and ω-HOG in the rest of the paper. Although the better performance of ω-Trajdesc versus the original Trajectory descriptor was expected, the one 222555555866 2. Trajectories obtained from optical and compensated flows. The green tail is the trajectory the current frame. The trajectories are sub-sampled for the sake of clarity. The frames are extracted Figure over every 15 frames with red dot indicating 5 frames in this example. DescriptorHollywood2HMDB51 BaseTrliaωnje- Tc(rtoarejrdpyreos[c2d9u]ced)54 7 1. 7 4% %2382.–89% BaseliHnωOe- (GHreOp [2rG9od]uced)4 451 . 658%%%2296.– 13%% Table 1. ω-Trajdesc and ω-HOG: Impact of compensating flow on Trajectory descriptor and HOG descriptors. achieved by ω-HOG might be surprising. Our interpretation is that HOG captures more context with the modified trajectories. More precisely, the original HOG descriptor is computed from a 2D+t sub-volume aligned with the corresponding trajectory and hence represents the appearance along the trajectory shape. When using ω-flow, we do not align the video sequence. As a result, the ω-HOG descriptor is no more computed around the very same tracked physical point in the space-time volume but around points lying in a patch of the initial feature point, whose size depends on the affine flow magnitude. ω-HOG can be viewed as a “patchbased” computation capturing more information about the appearance of the background or of the moving foreground. As for ω-trajectories, they are closer to the real trajectories of the moving actors as they usually cancel the camera movement, and so, more easier to train and recognize. Impact on HOF. The ω-flow impacts computation used as an input to HOF computation itself. Therefore, HOF can both types of trajectories (ω-trajectories both the trajectory and the descriptor be computed along or those extracted MethodHollywood2HMDB51 Table(ω2rHf.-alocO IwomkF)inpHgacOtFobf[2u9ωhsb]i:f-nlo ωgwot-hwωHOflFown5H 02 34O. 58291F% %descripto3 r706s38.:–1076% m%APfor Hollywood2 and average accuracy for HMDB5 1. The ω-HOF is used in subsequent evaluations. from flow) and can encode both kinds of flows (ω-flow or flow). For the sake of completeness, we evaluate all the variants as well as the combination of both flows in the descriptor computation stage. The results are presented in Table 2 and demonstrate the significant improvement obtained by computing the HOF descriptor with the ω-flow instead of the optical flow. Note that the type of trajectories which is used, either “Tracking flow” or “Tracking ω-flow”, has a limited impact in this case. From now on, we only consider the “Tracking ω-flow” case where HOF is computed along ω-trajectories. Interestingly, combining the HOF computed from the flow and the ω-flow further improves the results. This suggests that the two flow fields are complementary and the affine flow that was subtracted from ω-flow brings in additional information. For the sake of brevity, the combination of the two kinds of HOF, i.e., computed from the flow and the ω-flow using ω-trajectories, is referred to as the ω-HOF 222555555977 MethodHollywood2HMDB51 Tab(lerT3a.cIkmM inpBgMacHgωtBf-loH w [u2)s9in]gω f-lo wo MBH5 d42 e.052s7c% riptos:m34A90P.–3769f% orHllywood2 and average accuracy for HMDB5 1. DTerHasMjcBrOeblitpHGeFor4.ySumTωraw- frcilykto hw ionfgtheduωpCs-fcaolrtωmeiwNp-df/tl+Aωoutrw-finlogwthdesωcr- isTpc-fHtrMloaiOjrBpdswtGeHFosrc descriptor in the rest of this paper. Compared to the HOF baseline, the ω-HOF descriptor achieves a gain of +3.1% of mAP on Hollywood 2 and of +7.8% on HMDB51. Impact on MBH. Since MBH is computed from gradient of flow and cancel the constant motion, there is practically no benefit in using the ω-flow to compute the MBH descriptors, as shown in Table 3. However, by tracking ω-flow, the performance improves by around 1.3% for HMDB5 1 dataset and drops by around 1.5% for Hollywood2. This relative performance depends on the encoding technique. We will come back on this descriptor when considering another encoding scheme for local descriptors in Section 6. 4.3. Summary of compensated descriptors Table 4 summarizes the refined versions of the descriptors obtained by exploiting the ω-flow, and both ω-flow and the optical flow in the case of HOF. The revisited descriptors considerably improve the results compared to the orig- inal ones, with the noticeable exception of ω-MBH which gives mixed performance with a bag-of-features encoding scheme. But we already mention as this point that this incongruous behavior of ω-MBH is stabilized with the VLAD encoding scheme considered in Section 6. Another advantage of tracking the compensated flow is that fewer trajectories are produced. For instance, the total number of trajectories decreases by about 9. 16% and 22.81% on the Hollywood2 and HMDB51 datasets, respectively. Note that exploiting both the flow and the ω-flow do not induce much computational overhead, as the latter is obtained from the flow and the affine flow which is computed in real-time and already used to get the ω-trajectories. The only additional computational cost that we introduce by using the descriptors summarized in Table 4 is the computation of a second HOF descriptor, but this stage is relatively efficient and not the bottleneck of the extraction procedure. 5. Divergence-Curl-Shear descriptor This section introduces a new descriptor encoding the kinematic properties of motion discussed in Section 2.2. It is denoted by DCS in the rest of this paper. Combining kinematic features. The spatial derivatives are computed for the horizontal and vertical components of the flow field, which are used in turn to compute the divergence, curl and shear scalar values, see Equation 3. We consider all possible pairs of kinematic features, namely (div, curl), (div, shear) and (curl, shear). At each × ×× pixel, we compute the orientation and magnitude of the 2-D vector corresponding to each of these pairs. The orientation is quantized into histograms and the magnitude is used for weighting, similar to SIFT. Our motivation for encoding pairs is that the joint distribution of kinematic features conveys more information than exploiting them independently. Implementation details. The descriptor computation and parameters are similar to HOG and other popular descriptors such as MBH, HOF. We obtain 8-bin histograms for each of the three feature pairs or components of DCS. The range of possible angles is 2π for the (div,curl) pair and π for the other pairs, because the shear is always positive. The DCS descriptor is computed for a space-time volume aligned with a trajectory, as done with the four descriptors mentioned in the previous section. In order to capture the spatio-temporal structure of kinematic features, the volume (32 32 pixels and L = 15 frames) is subdivided into a spatio-temporal grid nofd s Lize = nx 5× f ny m×e nt, sw situhb nx =de ny =to 2a and nt = 3. These parameters ×hnave× × bneen fixed for the sake of consistency with the other descriptors. For each pair of kinematic features, each cell in the grid is represented by a histogram. The resulting local descriptors have a dimensionality equal to 288 = nx ny nt 8 3. At the video level, these descriptors are nenc×od end i×nto 8 a single vector representation using either BOF or the VLAD encoding scheme introduced in the next section. 6. VLAD in actions VLAD [8] is a descriptor encoding technique that aggregates the descriptors based on a locality criterion in the feature space. To our knowledge, this technique has never been considered for action recognition. Below, we briefly introduce this approach and give the performance achieved for all the descriptors introduced along the previous sections. VLAD in brief. Similar to BOF, VLAD relies on a codebook C = {c1, c2 , ...ck} of k centroids learned by k-means. bTohoek representation is ob}t oaifn ked c by summing, efodr b yea kch-m mveiasunasl. word ci, the differences x − ci of the vectors x assigned to ci, thereby producing a sv exct −or c representation oflength d×k, 222555556088 DMeBscHriptorV5 LH.A1o%Dlywo5Bo4d.O2 %F4V3L.3HA%MD B35B91.O7%F Taωbl-eDHM5rOCBa.FSjGPdHe+rsωfco-mMHaBOnFeofV54L2936A.51D% with5431ω208-.5T96% rajde3s42c97158,.ω3% -HOG342,58019ω.6-% HOF descriptors and their combination. where d is the dimension ofthe local descriptors. We use the codebook size, k = 256. Despite this large dimensionality, VLAD is efficient because it is effectively compared with a linear kernel. VLAD is post-processed using a componentwise power normalization, which dramatically improves its performance [8]. While cross validating the parameter α involved in this power normalization, we consistently observe, for all the descriptors, a value between 0.15 and 0.3. Therefore, this parameter is set to α = 0.2 in all our experiments. For classification, we use a linear SVM and oneagainst-rest approach everywhere, unless stated otherwise. Impact on existing descriptors. We employ VLAD because it is less sensitive to quantization parameters and appears to provide better performance with descriptors having a large dimensionality. These properties are interesting in our case, because the quantization parameters involved in the DCS and MBH descriptors have been used unchanged in Section 4 for the sake of direct comparison. They might be suboptimal when using the ω-flow instead of the optical flow on which they have initially been optimized [29]. Results for MBH and ω-MBH in Table 5 supports this argument. When using VLAD instead of BOF, the scores are stable in both the cases and there is no mixed inference as that observed in Table 3. VLAD also has significant positive influence on accuracy of ω-DCS descriptor. We also observe that ω-DCS is complementary to ω-MBH and adds to the performance. Still DCS is probably not best utilized in the current setting of parameters. In case of ω-Trajdesc and ω-HOG, the scores are better with BOF on both the datasets. ω-HOF with VLAD improves on HMDB5 1, but remains equivalent for Hollywood2. Although BOF leads to better scores for the descriptors considered individually, their combination with VLAD outperforms the BOF. 7. Comparison with the state of the art This section reports our results with all descriptors combined and compares our method with the state of the art. TrajectorCy+omHbOiGna+tHioOnF+ MDBCHSHol5 l98y.w76%o%od2H4M489.D02%B%51 All ω-descriptors all five compensated descriptors using combined62.5%52.1% Table 6. Combination of VLAD representation. WU*JVliaOnughreM tHaeolth. [yo2w9d87o] 256 0985. 37% SKa*duOeJinhrau tnegdMteHatlMa.h [ol1Dd.0B[91]25 24 609.8172% Table 7. Comparison with the state of the art on Hollywood2 and HMDB5 1 datasets. *Vig et al. [28] gets 61.9% by using external eye movements data. *Jiang et al. [9] used one-vs-one multi class SVM while our and other methods use one-vs-rest SVMs. With one-against-one multi class SVM we obtain 45. 1% for HMDB51. Descriptor combination. Table 6 reports the results obtained when the descriptors are combined. Since we use VLAD, our baseline is updated that is combination of Trajectory, HOG, HOF and MBH with VLAD representation. When DCS is added to the baseline there is an improvement of 0.9% and 1.2%. With combination of all five compensated descriptors we obtain 62.5% and 52.1% on the two datasets. This is a large improvement even over the updated baseline, which shows that the proposed motion compensation and the way we exploit it are significantly important for action recognition. The comparison with the state of the art is shown in Table 7. Our method outperforms all the previously reported results in the literature. In particular, on the HMDB51 dataset, the improvement over the best reported results to date is more than 11% in average accuracy. Jiang el al. [9] used a one-against-one multi-class SVM, which might have resulted in inferior scores. With a similar multi-class SVM approach, our method obtains 45. 1%, which remains significantly better than their result. All others results were reported with one-against-rest approach. On Olympic Sports dataset we obtain mAP of 83.2% with ‘All ω-descriptors combined’ and the improvement is mostly because of VLAD and ω-flow. The best reported mAPs on this dataset are Liu et al. [14] (74.4%) and Jiang et al. [9] (80.6%), which we exceed convincingly. Gaidon et al. [6] reports the best average accuracy of 82.7%. 8. Conclusions This paper first demonstrates the interest of canceling the dominant motion (predominantly camera motion) to make the visual motion truly related to actions, for both the trajectory extraction and descriptor computation stages. It pro222555556199 duces significantly better versions (called compensated descriptors) of several state-of-the-art local descriptors for action recognition. The simplicity, efficiency and effectiveness of this motion compensation approach make it applicable to any action recognition framework based on motion descriptors and trajectories. The second contribution is the new DCS descriptor derived from the first-order scalar motion quantities specifying the local motion patterns. It captures additional information which is proved complementary to the other descriptors. Finally, we show that VLAD encoding technique instead of bag-of-words boosts several action descriptors, and overall exhibits a significantly better performance when combining different types of descriptors. Our contributions are all complementary and significantly outperform the state of the art when combined, as demon- strated by our extensive experiments on the Hollywood 2, HMDB51 and Olympic Sports datasets. Acknowledgments This work was supported by the Quaero project, funded by Oseo, French agency for innovation. We acknowledge Heng Wang’s help for reproducing some of their results. References [1] S. Ali and M. Shah. Human action recognition in videos using kinematic features and multiple instance learning. IEEE T-PAMI, 32(2):288–303, Feb. 2010. [2] T. Brox and J. Malik. Object segmentation by long term analysis of point trajectories. In ECCV, Sep. 2010. [3] N. Dalal and B. Triggs. Histograms of oriented gradients for human detection. In CVPR, Jun. 2005. [4] N. Dalal, B. Triggs, and C. Schmid. Human detection using oriented histograms of flow and appearance. In ECCV, May 2006. [5] P. Dollar, V. Rabaud, G. Cottrell, and S. Belongie. Behavior recognition via sparse spatio-temporal features. In VS-PETS, Oct. 2005. [6] A. Gaidon, Z. Harchaoui, and C. Schmid. Recognizing activities with cluster-trees of tracklets. In BMVC, Sep. 2012. [7] A. Hervieu, P. Bouthemy, and J.-P. Le Cadre. A statistical video content recognition method using invariant features on object trajectories. IEEE T-CSVT, 18(1 1): 1533–1543, 2008. [8] H. J ´egou, F. Perronnin, M. Douze, J. S ´anchez, P. P ´erez, and C. Schmid. Aggregating local descriptors into compact [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] codes. IEEE T-PAMI, 34(9):1704–1716, 2012. Y.-G. Jiang, Q. Dai, X. Xue, W. Liu, and C.-W. Ngo. Trajectory-based modeling of human actions with motion reference points. In ECCV, Oct. 2012. O. Kliper-Gross, Y. Gurovich, T. Hassner, and L. Wolf. Motion interchange patterns for action recognition in unconstrained videos. In ECCV, Oct. 2012. H. Kuehne, H. Jhuang, E. Garrote, T. Poggio, and T. Serre. Hmdb: A large video database for human motion recognition. In ICCV, Nov. 2011. I. Laptev and T. Lindeberg. Space-time interest points. In ICCV, Oct. 2003. I. Laptev, M. Marzalek, C. Schmid, and B. Rozenfeld. Learning realistic human actions from movies. In CVPR, Jun. 2008. J. Liu, B. Kuipers, and S. Savarese. Recognizing human actions by attributes. In CVPR, Jun. 2011. D. Lowe. Distinctive image features from scale-invariant keypoints. IJCV, 60(2):91–1 10, Nov. 2004. M. Marzalek, I. Laptev, and C. Schmid. Actions in context. In CVPR, Jun. 2009. P. Matikainen, M. Hebert, and R. Sukthankar. Trajectons: Action recognition through the motion analysis of tracked features. In Workshop on Video-Oriented Object and Event Classification, ICCV, Sep. 2009. R. Messing, C. J. Pal, and H. A. Kautz. Activity recognition using the velocity histories of tracked keypoints. In ICCV, Sep. 2009. J. C. Niebles, C.-W. Chen, and F.-F. Li. Modeling temporal structure of decomposable motion segments for activity classification. In ECCV, Sep. 2010. [20] J.-M. Odobez and P. Bouthemy. Robust multiresolution estimation of parametric motion models. Jal of Vis. Comm. and Image Representation, 6(4):348–365, Dec. 1995. [21] G. Piriou, P. Bouthemy, and J.-F. Yao. Recognition of dynamic video contents with global probabilistic models of visual motion. IEEE T-IP, 15(1 1):3417–3430, 2006. [22] S. Sadanand and J. J. Corso. Action bank: A high-level representation of activity in video. In CVPR, Jun. 2012. [23] C. Schmid and R. Mohr. Local grayvalue invariants for image retrieval. IEEE T-PAMI, 19(5):530–534, May 1997. [24] J. Sivic and A. Zisserman. Video Google: A text retrieval approach to object matching in videos. In ICCV, pages 1470– 1477, Oct. 2003. [25] J. Sun, X. Wu, S. Yan, L. F. Cheong, T.-S. Chua, and J. Li. Hierarchical spatio-temporal context modeling for action recognition. In CVPR, Jun. 2009. [26] H. Uemura, S. Ishikawa, and K. Mikolajczyk. Feature tracking and motion compensation for action recognition. In BMVC, Sep. 2008. [27] M. M. Ullah, S. N. Parizi, and I. Laptev. Improving bag-offeatures action recognition with non-local cues. In BMVC, Sep. 2010. [28] E. Vig, M. Dorr, and D. Cox. Saliency-based space-variant descriptor sampling for action recognition. In ECCV, Oct. 2012. [29] H. Wang, A. Kl¨ aser, C. Schmid, and C.-L. Liu. Action recognition by dense trajectories. In CVPR, Jun. 2011. [30] H. Wang, M. M. Ullah, A. Kl¨ aser, I. Laptev, and C. Schmid. Evaluation of local spatio-temporal features for action recognition. In BMVC, Sep. 2009. [3 1] G. Willems, T. Tuytelaars, and L. J. V. Gool. An efficient dense and scale-invariant spatio-temporal interest point detector. In ECCV, Oct. 2008. [32] S. Wu, O. Oreifej, and M. Shah. Action recognition in videos acquired by a moving camera using motion decomposition of lagrangian particle trajectories. In ICCV, Nov. 2011. 222555666200

5 0.73911554 124 cvpr-2013-Determining Motion Directly from Normal Flows Upon the Use of a Spherical Eye Platform

Author: Tak-Wai Hui, Ronald Chung

Abstract: We address the problem of recovering camera motion from video data, which does not require the establishment of feature correspondences or computation of optical flows but from normal flows directly. We have designed an imaging system that has a wide field of view by fixating a number of cameras together to form an approximate spherical eye. With a substantially widened visual field, we discover that estimating the directions of translation and rotation components of the motion separately are possible and particularly efficient. In addition, the inherent ambiguities between translation and rotation also disappear. Magnitude of rotation is recovered subsequently. Experimental results on synthetic and real image data are provided. The results show that not only the accuracy of motion estimation is comparable to those of the state-of-the-art methods that require explicit feature correspondences or optical flows, but also a faster computation time.

6 0.69347161 118 cvpr-2013-Detecting Pulse from Head Motions in Video

7 0.68491375 88 cvpr-2013-Compressible Motion Fields

8 0.67524385 334 cvpr-2013-Pose from Flow and Flow from Pose

9 0.6652773 10 cvpr-2013-A Fully-Connected Layered Model of Foreground and Background Flow

10 0.64025843 333 cvpr-2013-Plane-Based Content Preserving Warps for Video Stabilization

11 0.63607717 203 cvpr-2013-Hierarchical Video Representation with Trajectory Binary Partition Tree

12 0.61743504 37 cvpr-2013-Adherent Raindrop Detection and Removal in Video

13 0.60944825 313 cvpr-2013-Online Dominant and Anomalous Behavior Detection in Videos

14 0.60649937 46 cvpr-2013-Articulated and Restricted Motion Subspaces and Their Signatures

15 0.5963912 167 cvpr-2013-Fast Multiple-Part Based Object Detection Using KD-Ferns

16 0.59085298 363 cvpr-2013-Robust Multi-resolution Pedestrian Detection in Traffic Scenes

17 0.58640587 398 cvpr-2013-Single-Pedestrian Detection Aided by Multi-pedestrian Detection

18 0.58509403 272 cvpr-2013-Long-Term Occupancy Analysis Using Graph-Based Optimisation in Thermal Imagery

19 0.58356512 345 cvpr-2013-Real-Time Model-Based Rigid Object Pose Estimation and Tracking Combining Dense and Sparse Visual Cues

20 0.58181369 383 cvpr-2013-Seeking the Strongest Rigid Detector


similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(10, 0.104), (16, 0.035), (26, 0.058), (28, 0.02), (33, 0.319), (38, 0.158), (67, 0.112), (69, 0.041), (80, 0.012), (87, 0.07)]

similar papers list:

simIndex simValue paperId paperTitle

1 0.9598754 350 cvpr-2013-Reconstructing Loopy Curvilinear Structures Using Integer Programming

Author: Engin Türetken, Fethallah Benmansour, Bjoern Andres, Hanspeter Pfister, Pascal Fua

Abstract: We propose a novel approach to automated delineation of linear structures that form complex and potentially loopy networks. This is in contrast to earlier approaches that usually assume a tree topology for the networks. At the heart of our method is an Integer Programming formulation that allows us to find the global optimum of an objective function designed to allow cycles but penalize spurious junctions and early terminations. We demonstrate that it outperforms state-of-the-art techniques on a wide range of datasets.

2 0.93852228 263 cvpr-2013-Learning the Change for Automatic Image Cropping

Author: Jianzhou Yan, Stephen Lin, Sing Bing Kang, Xiaoou Tang

Abstract: Image cropping is a common operation used to improve the visual quality of photographs. In this paper, we present an automatic cropping technique that accounts for the two primary considerations of people when they crop: removal of distracting content, and enhancement of overall composition. Our approach utilizes a large training set consisting of photos before and after cropping by expert photographers to learn how to evaluate these two factors in a crop. In contrast to the many methods that exist for general assessment of image quality, ours specifically examines differences between the original and cropped photo in solving for the crop parameters. To this end, several novel image features are proposed to model the changes in image content and composition when a crop is applied. Our experiments demonstrate improvements of our method over recent cropping algorithms on a broad range of images.

3 0.93071783 78 cvpr-2013-Capturing Layers in Image Collections with Componential Models: From the Layered Epitome to the Componential Counting Grid

Author: Alessandro Perina, Nebojsa Jojic

Abstract: Recently, the Counting Grid (CG) model [5] was developed to represent each input image as a point in a large grid of feature counts. This latent point is a corner of a window of grid points which are all uniformly combined to match the (normalized) feature counts in the image. Being a bag of word model with spatial layout in the latent space, the CG model has superior handling of field of view changes in comparison to other bag of word models, but with the price of being essentially a mixture, mapping each scene to a single window in the grid. In this paper we introduce a family of componential models, dubbed the Componential Counting Grid, whose members represent each input image by multiple latent locations, rather than just one. In this way, we make a substantially more flexible admixture model which captures layers or parts of images and maps them to separate windows in a Counting Grid. We tested the models on scene and place classification where their com- ponential nature helped to extract objects, to capture parallax effects, thus better fitting the data and outperforming Counting Grids and Latent Dirichlet Allocation, especially on sequences taken with wearable cameras.

same-paper 4 0.92405206 158 cvpr-2013-Exploring Weak Stabilization for Motion Feature Extraction

Author: Dennis Park, C. Lawrence Zitnick, Deva Ramanan, Piotr Dollár

Abstract: We describe novel but simple motion features for the problem of detecting objects in video sequences. Previous approaches either compute optical flow or temporal differences on video frame pairs with various assumptions about stabilization. We describe a combined approach that uses coarse-scale flow and fine-scale temporal difference features. Our approach performs weak motion stabilization by factoring out camera motion and coarse object motion while preserving nonrigid motions that serve as useful cues for recognition. We show results for pedestrian detection and human pose estimation in video sequences, achieving state-of-the-art results in both. In particular, given a fixed detection rate our method achieves a five-fold reduction in false positives over prior art on the Caltech Pedestrian benchmark. Finally, we perform extensive diagnostic experiments to reveal what aspects of our system are crucial for good performance. Proper stabilization, long time-scale features, and proper normalization are all critical.

5 0.91804129 233 cvpr-2013-Joint Sparsity-Based Representation and Analysis of Unconstrained Activities

Author: Raghuraman Gopalan

Abstract: While the notion of joint sparsity in understanding common and innovative components of a multi-receiver signal ensemble has been well studied, we investigate the utility of such joint sparse models in representing information contained in a single video signal. By decomposing the content of a video sequence into that observed by multiple spatially and/or temporally distributed receivers, we first recover a collection of common and innovative components pertaining to individual videos. We then present modeling strategies based on subspace-driven manifold metrics to characterize patterns among these components, across other videos in the system, to perform subsequent video analysis. We demonstrate the efficacy of our approach for activity classification and clustering by reporting competitive results on standard datasets such as, HMDB, UCF-50, Olympic Sports and KTH.

6 0.91515553 171 cvpr-2013-Fast Trust Region for Segmentation

7 0.91285157 119 cvpr-2013-Detecting and Aligning Faces by Image Retrieval

8 0.91226435 254 cvpr-2013-Learning SURF Cascade for Fast and Accurate Object Detection

9 0.91117603 322 cvpr-2013-PISA: Pixelwise Image Saliency by Aggregating Complementary Appearance Contrast Measures with Spatial Priors

10 0.90791529 339 cvpr-2013-Probabilistic Graphlet Cut: Exploiting Spatial Structure Cue for Weakly Supervised Image Segmentation

11 0.90723616 122 cvpr-2013-Detection Evolution with Multi-order Contextual Co-occurrence

12 0.90669656 94 cvpr-2013-Context-Aware Modeling and Recognition of Activities in Video

13 0.90651405 438 cvpr-2013-Towards Pose Robust Face Recognition

14 0.90613252 4 cvpr-2013-3D Visual Proxemics: Recognizing Human Interactions in 3D from a Single Image

15 0.905963 202 cvpr-2013-Hierarchical Saliency Detection

16 0.90581769 204 cvpr-2013-Histograms of Sparse Codes for Object Detection

17 0.90561581 167 cvpr-2013-Fast Multiple-Part Based Object Detection Using KD-Ferns

18 0.9054029 104 cvpr-2013-Deep Convolutional Network Cascade for Facial Point Detection

19 0.90504128 416 cvpr-2013-Studying Relationships between Human Gaze, Description, and Computer Vision

20 0.90458012 338 cvpr-2013-Probabilistic Elastic Matching for Pose Variant Face Verification