Unsupervised content-based indexing for sports video retrieval

Abstract

This demonstration presents an interface to a corpus of broadcast baseball games that have been indexed using an unsupervised content-based method introduced here. The method uses the concept of a grounded language model to motivate a framework in which video is searched using natural language with no reliance on predetermined concepts or hand labeled events. The interface demonstrates the effectiveness of the technique and the ease of use it affords the user.

Using closed captions to train activity recognizers that improve video retrieval

“…Fleischman and Roy [8] used both captions and motion descriptions for baseball video to retrieve relevant clips given a textual query. Additionally, Fleischman and Roy [9] presented a method for using speech recognition on the soundtrack to further improve retrieval. They used an unsupervised Author Topic Model, a generalization of Latent Dirichlet Allocation, to learn correlations between caption text and encoded event representations.…”

“…Recently, significant progress has been made on activity recognition systems that detect specific human actions in real-world videos [6,15]. One application of recent interest is retrieving clips of particular events in sports videos such as baseball broadcasts [9]. Activity recognition in sports videos is particularly difficult because of the ambiguous video cues, background clutter, rapid change of actions, change in camera zoom and angle etc.…”

Detection of anomalous events in a tennis game using multimodal information

“…It is interesting both because of its rich audio-visual information content, which also has a strong inherent syntax, and because there are several useful practical applications of such analysis, such as highlight extraction [1], tactics analysis [2], computer-assisted refereeing [3]. Research in this area has also been influential in information retrieval [4], audio contents analysis [5], and tracking motion objects [6].…”

Enhancing sketch-based sport video retrieval by suggesting relevant motion paths

“…Request permissions from permissions@acm.org. SIGIR'14, July [6][7][8][9][10][11]2014 that is obtained by extracting player and ball movements either manually such as in [1], from broadcast videos such as in [2,3], or from on-the-field cameras specifically deployed to assist in providing tracking information such as in [4,5,6,7,8,9]. Recently, the emergence of light-weight wireless sensor devices explicitly designed for the sports domain [10,11] allows to capture a wider array of data including physiological data and, at the same time, obtain more accurate tracking information.…”