Saliency-based object recognition in video
Year: 2013
Type: preprint
Abstract: In this paper we study the problem of object recognition in egocentric video recorded with cameras worn by persons. This task hasgained much attention during the last years, since it has turned tobe a main building block for action recognition systems in applications involving wearable cameras, such as tele-medicine or lifelogging. Under these scenarios, an action can be effectively definedas a sequence of manipulated or observed objects, so that recognition becomes a relevant stage of the system. Furthermore, videosummarization tasks on such content is also driven by appearanceof semantic objects in camera field of view.One of the particularities of first-person camera videos is thatthey usually present a strong differentiation between active (manipulated or observed by the user wearing the camera) and passiveobjects (associated to background). In addition, spatial, temporaland geometric cues can be found in the video content that may helpto identify the active elements in the scene. These saliency featuresare related to the modelling of Human Visual System, but also tomotor coordination of eye, hand and body movements. In this paper, we discuss the automatic generation of saliency maps in video,and introduce a method that extends the well-known Bag-of-Words(BoW) paradigm with saliency information. We have assessed ourproposal in several egocentric video datasets, demonstrating that itnot only improves the BoW reference, but also achieves state-ofthe-art performance of e.g. part - based models, with noticeablylower computational times. The approach has tremendous perspectives for other User Generated mobile Content.
Source: HAL
Institution: Laboratoire Bordelais de Recherche en Informatique
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2610537166