Human Action Recognition Using Temporal Segmentation and Accordion Representation

Spatio-temporal pyramidal accordion representation for human action recognition

“…In this work, our motion descriptor achieves 57.5%. It outperforms the approaches proposed in [25,26,11,14] and gives similar results with the MBH+STP descriptor [6].…”

“…The MBH+STP descriptor [6] achieves 74.9%. In our previous work [14] we obtained a mAP equal to 72.5%. As shown in table 2 our descriptor (mAP=75.6%) outperforms MBH+STP descriptor [6] as well as all methods proposed in [4,24,25,14] and even in some cases by a significant margin.…”

“…The method gives a mAP equal to 55.7%. In our previous work [14], based on temporal segmentation, the Accordion representation is applied separately on each elementary motion segment. We obtained a mAP equal to 55.9%.…”

“…Recently, Wang et al [6] propose to use Spatio-Temporal Pyramid with the dense trajectory application and combination of different descriptors. In our previous work [14,13], the presented descriptor is based on the Accordion representation that transforms the video into a plan in order to put pixels with temporal adjacency in a spatial neighbourhood. Then, the Accordion representation is applied separately on each elementary motion segment.…”

Human action recognition based on multi-layer Fisher vector encoding method

“…Having no previous knowledge about the location of the person in each video frame, the human action in a video stream can be recovered from a great number of local descriptors extracted from the video frames (Sekma et al, 2013), (Dammak et al, 2012), , (Sekma et al, 2014). Local descriptors, coupled with the bag-of-words (BOW) encoding method (Sivic and Zisserman, 2003) (Mejdoub et al, 2008) (Mejdoub et al, 2007) have recently become a very popular video representation (Ben Aoun et al, 2014), (Knopp et al, 2010), (Laptev et al, 2008), (Wang et al, 2009), (Alexander et al, 2008), (Wang et al, 2011), (Raptis and Soatto, 2010), (Pyry et al, 2010), (Jiang et al, 2012) and (Jain et al, 2013).…”