Learning Discriminative Space–Time Action Parts from Weakly Labelled Videos

Sapienza, Michael; Cuzzolin, Fabio; Torr, Philip H. S.

doi:10.1007/s11263-013-0662-8

Cited by 54 publications

(23 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can be seen that the two-layer model had better performance than the low-level encoding using different descriptors. In addition, we compare the proposed method with a few classic methods on KTH, J-HMDB, and YouTube datasets, such as DT + BoVW [1], mid-level parts [21], traditional FV [17], stacked FV [17], DT + BOW [10], and IDT + FV [17]. As shown in Table 2, the twolayer model obtains 67.4% and 87.6% accuracy on J-HMDB and YouTube datasets, respectively.…”

Section: Experiments Resultmentioning

confidence: 99%

“…YouTube J-HMDB ISA [15] 86.5% Liu et al [16] 71.2% Traditional FV [17] 62.83% Yeffet and Wolf [18] 90.1% Ikizler-Cinbis and Sclaroff [19] 75.21% Stacked FV [17] 59.27% Cheng et al [20] 89.7% DT + BoVW [1] 85.4% DT + BOW [10] 56.6% Le et al [15] 93.9% Mid-level parts [21] 84.5% IDT + FV [17] 62.8% Two-layer model 92.6% Two-layer model 87.6% Two-layer model 67.4%…”

Section: Kthmentioning

confidence: 99%

See 1 more Smart Citation

Representation for Action Recognition Using Trajectory-Based Low-Level Local Feature and Mid-Level Motion Feature

Wang

Zhang

2017

Applied Computational Intelligence and Soft Computing

View full text Add to dashboard Cite

The dense trajectories and low-level local features are widely used in action recognition recently. However, most of these methods ignore the motion part of action which is the key factor to distinguish the different human action. This paper proposes a new two-layer model of representation for action recognition by describing the video with low-level features and mid-level motion part model. Firstly, we encode the compensated flow ( -flow) trajectory-based local features with Fisher Vector (FV) to retain the low-level characteristic of motion. Then, the motion parts are extracted by clustering the similar trajectories with spatiotemporal distance between trajectories. Finally the representation for action video is the concatenation of low-level descriptors encoding vector and motion part encoding vector. It is used as input to the LibSVM for action recognition. The experiment results demonstrate the improvements on J-HMDB and YouTube datasets, which obtain 67.4% and 87.6%, respectively.

show abstract

Section: Experiments Resultmentioning

confidence: 99%

Section: Kthmentioning

confidence: 99%

Representation for Action Recognition Using Trajectory-Based Low-Level Local Feature and Mid-Level Motion Feature

Wang

Zhang

2017

Applied Computational Intelligence and Soft Computing

View full text Add to dashboard Cite

show abstract

“…We show the classification accuracy in Table 3. [3] is the Table 3 Comparison of our proposed S-T Saliency to the state-of-the-art on both datasets Algorithms UCF11 HMDB51 HoG [3] 74.5 40.2 HoF [3] 72.8 48.9 MBH [3] 83.9 52.1 DT [3] 84.2 54.7 iDT [18], [19] 90.7 57.2 Mid-level parts [20] 84.5 37.2 CompactFV [5] 89 54.8 Our best 90.6 60. 1 original dense trajectory, [18] is the improved dense trajectories which compensated the camera motion by using RANSAC [21] to compute homography and remove the camera motion in each frame, they also use fisher vector [22] for feature encoding, Oneata et.al.…”

Section: Comparison With State-of-the-artmentioning

confidence: 99%

“…[5] extract spatial fisher vectors based on MBH and SIFT descriptors. [20] uses a deformable model to learn space-time parts.…”

Section: Comparison With State-of-the-artmentioning

confidence: 99%

Mining Spatial Temporal Saliency Structure for Action Recognition

Liu

et al. 2016

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYTraditional action recognition approaches use pre-defined rigid areas to process the space-time information, e.g. spatial pyramids, cuboids. However, most action categories happen in an unconstrained manner, that is, the same action in different videos can happen at different places. Thus we need a better video representation to deal with the space-time variations. In this paper, we introduce the idea of mining spatial temporal saliency. To better handle the uniqueness of each video, we use a space-time over-segmentation approach, e.g. supervoxel. We choose three different saliency measures that take not only the appearance cues, but also the motion cues into consideration. Furthermore, we design a category-specific mining process to find the discriminative power in each action category. Experiments on action recognition datasets such as UCF11 and HMDB51 show that the proposed spatial temporal saliency video representation can match or surpass some of the state-of-the-art alternatives in the task of action recognition.

show abstract

“…Besides other domains such as scene, object recognition and tracking [26,5], MIL has been used in the categorization of singleton human actions in [2,22,43]. Prabhakar and Rehg [36] use multiple instance learning to infer the labels of causal sets which temporally co-occur in turn-taking interactions.…”

Section: Related Workmentioning

confidence: 99%

Two-person interaction recognition via spatial multiple instance embedding

Sener

Ikizler-Cinbis

2015

Journal of Visual Communication and Image Representation

View full text Add to dashboard Cite

a b s t r a c tIn this work, we look into the problem of recognizing two-person interactions in videos. Our method integrates multiple visual features in a weakly supervised manner by utilizing an embedding-based multiple instance learning framework. In our proposed method, first, several visual features that capture the shape and motion of the interacting people are extracted from each detected person region in a video. Then, twoperson visual descriptors are formed. Since the relative spatial locations of interacting people are likely to complement the visual descriptors, we propose to use spatial multiple instance embedding, which implicitly incorporates the distances between people into the multiple instance learning process. Experimental results on two benchmark datasets validate that using two-person visual descriptors together with spatial multiple instance learning offers an effective way for inferring the type of the interaction.

show abstract

Learning Discriminative Space–Time Action Parts from Weakly Labelled Videos

Cited by 54 publications

References 40 publications

Representation for Action Recognition Using Trajectory-Based Low-Level Local Feature and Mid-Level Motion Feature

Representation for Action Recognition Using Trajectory-Based Low-Level Local Feature and Mid-Level Motion Feature

Mining Spatial Temporal Saliency Structure for Action Recognition

Two-person interaction recognition via spatial multiple instance embedding

Contact Info

Product

Resources

About