Motion Trend Patterns for Action Modelling and Recognition

Nguyễn, Thanh Phương; Manzanera, Antoine; Garrigues, Matthieu

doi:10.1007/978-3-642-40261-6_43

Cited by 2 publications

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On KTH (6 actions) we just obtain 91.1% recognition accuracy, which is not best but comparable to 91.7% in [27], 94.5% in [28] and 92.5% in [29]. It should be noticed that on this database those methods use a complex classifier or multiple descriptors combined to obtain better performance.…”

Section: Comparisons With Representative Algorithmsmentioning

confidence: 78%

See 1 more Smart Citation

Bio-inspired model with dual visual pathways for human action recognition

Cai

Qing

2014

2014 9th International Symposium on Communication Systems, Networks &Amp; Digital Sign (CSNDSP)

View full text Add to dashboard Cite

Spatio-temporal interesting points feature is a key technology for a wide class of computer vision approaches to recognize human actions. In this paper, a novel bio-inspired model based spatio-temporal interesting points (BIM-STIP) framework is proposed. Different from traditional STIP framework, the introduction of bio-inspired model provides a biological theory for interest points detection and spatio-temporal descriptor construction. Moreover, unlike existing bio-inspired model for action recognition with a single pathway, a dual pathway joint model with dorsal stream and ventral stream is constructed. Finally, this paper demonstrate how these feature to be used in a standard STIP framework for action recognition. Promising experimental results show that the proposed framework outperforms most of existing algorithms in action recognition, which encourages us to develop the BIM-STIP framework to other applications in future.

show abstract

Section: Comparisons With Representative Algorithmsmentioning

confidence: 78%

“…Unified [32] 87.3% 93.6% Speech [33] 90.3% 96.2% SMT [27] 91.7% 93.4% HOG-OF [28] 94.3% 93.6% MTP [29] 92 …”

Section: B Comparisons With Other Bio-inspired Methodsmentioning

confidence: 99%

Bio-inspired model with dual visual pathways for human action recognition

Cai

Qing

2014

2014 9th International Symposium on Communication Systems, Networks &Amp; Digital Sign (CSNDSP)

View full text Add to dashboard Cite

show abstract

“…Furthermore, it achieves a trade-off between dense optical flow and long term point tracking, which makes it a very versatile brick that can be used in many different applications: non rigid object tracking, structure from motion, video stabilisation, video segmentation, action recognition and so on. It has been used already to build action descriptors using beam of trajectories [10,9]. In [9], it is shown that the performance of action classification is higher when the number of trajectory increases, up to a certain limit: when the quality of matching begins to drop.…”

Section: Discussionmentioning

confidence: 99%

Video Extruder: a semi-dense point tracker for extracting beams of trajectories in real time

Garrigues

Manzanera

Berthet

2014

J Real-Time Image Proc

Self Cite

View full text Add to dashboard Cite

Two crucial aspects of visual point tracking are addressed in this paper. First, the algorithm should track as many points as possible reliably. Second, the computation should be fast enough, which is challenging on low power embedded platforms. We propose a new multi-scale semi dense point tracker called Video Extruder, whose purpose is to fill the gap between short term, dense motion estimation (optical flow) and long term, sparse salient point tracking. This paper presents a new detector, including a new salience function with low computational complexity and a new selection strategy that allows to obtain a large number of keypoints. Its density and reliability in mobile video scenarios are compared with those of FAST detector. Then, a multi-scale prediction and a matching strategy are presented, based on a hybrid regional coarse-to-fine and temporal prediction, which provides robustness to large camera and object accelerations. Filtering and merging strategies are then used to eliminate most of the wrong or useless trajectories. Thanks to its high degree of parallelism, the proposed algorithm extracts beams of trajectories from the video in a very fast way. We compare it with the state-of-the-art pyramidal Lucas Kanade point tracker and show that, in fast mobile video scenarios, it yields similar quality results, while being up to one order of magnitude faster. Three different parallel implementations of this tracker are presented, including multi-core CPU, GPU and ARM SoCs. On a commodity 2010 CPU, it can track 8 500 points in a 640 × 480 video at 150 Hz.

show abstract

Motion Trend Patterns for Action Modelling and Recognition

Cited by 2 publications

References 17 publications

Bio-inspired model with dual visual pathways for human action recognition

Bio-inspired model with dual visual pathways for human action recognition

Video Extruder: a semi-dense point tracker for extracting beams of trajectories in real time

Contact Info

Product

Resources

About