Instance search retrospective with focus on TRECVID

Abstract

This paper presents an overview of the Video Instance Search benchmark which was run over a period of 6 years (2010–2015) as part of the TREC Video Retrieval (TRECVID) workshop series. The main contributions of the paper include i) an examination of the evolving design of the evaluation framework and its components (system tasks, data, measures); ii) an analysis of the influence of topic characteristics (such as rigid/non rigid, planar/non-planar, stationary/mobile on performance; iii) a high-level overview of results and best-performing approaches. The Instance Search (INS) benchmark worked with a variety of large collections of data including Sound & Vision, Flickr, BBC (British Broadcasting Corporation) Rushes for the first 3 pilot years and with the small world of the BBC Eastenders series for the last 3 years.

[Retracted] Mean Shift Fusion Color Histogram Algorithm for Nonrigid Complex Target Tracking in Sports Video

“…rough the main operations of these two techniques, video image data are analyzed and expressed to locate the position of the human target and use the data for subsequent classification and recognition [11]. ere are many similarities between human segmentation and tracking techniques, one of which is that both methods operate with targets from video image capture, and both video images contain commonly desired target behavior segments; both techniques have high requirements in terms of real-time, accuracy, and robustness of image processing.…”

Deeply Activated Salient Region for Instance Search

“…Instance search was addressed as a sub-image retrieval task before CNNs were introduced [1] to visual object detection. The main image features being employed for this task are hand-crafted local descriptors such as SIFT and SURF.…”

“…When instance search was irst addressed in [1], the problem was coined as a sub-image retrieval task. Handcrafted features such as SIFT [22] and SURF [3] that are superior in local image matching were de-facto descriptors at that time.…”

“…Handcrafted features such as SIFT [22] and SURF [3] that are superior in local image matching were de-facto descriptors at that time. Although encouraging results are reported [1], these approaches are known to be limited to match textureless image patches and instances undergone non-rigid motions. While most of the descriptors are capable of generating thousands of local features from an image for matching, these features are extracted from regions rich of textures or corners.…”

Modelling perceptions on the evaluation of video summarization

“…Video summarization techniques create automatic video summaries by meeting three requirements: The presence of relevant video entities and events, elimination of redundant information, and generation of as much useful information as possible (Truong & Venkatesh, 2007). Truong & Venkatesh (2007) describe some video summarization applications such as browsing and retrieval, which is responsible for assisting users on searching and browsing tasks (Awad et al, 2017b;Arman et al, 1994;Zhang et al, 1997;Haojin Yang & Meinel, 2014), computational reduction and content analysis, used on semantic abstraction of information to reduce the computational complexity (Plummer et al, 2017), story navigation and video editing, which help users on navigating through a video (Nguyen et al, 2012), and highlighting, targeted on detection of important events in videos (Yao et al, 2016;Gygli et al, 2014;Xiong et al, 2003). On each of these applications, video summarization techniques try to mimic the ways humans comprehend the most important parts of a video.…”