An efficient Two Pathways Deep Architecture for Soccer Goal Recognition towards Soccer Highlight Summarization
Subject Areas : AI and Robotics
Amirhosein Zangane
1
,
Mehdi Jampour
2
,
Kamran Layeghi
3
1 - Doctoral student of North Tehran branch
2 - Assistant Professor, Faculty of Computer Engineering
3 - دانشگاه آزاد
Keywords: Dual-Path Deep Learning Architecture, Feature Combination, Deep VGG, Classic Features, Common Architecture,
Abstract :
In this paper, an automated method has been presented using a dual-path deep learning architecture model for the problem of soccer video analysis and it emphasizes the gate recognition as one of the most important elements of the goal event that is the most important soccer game event. The proposed architecture is considered as an extended form of the VGG 13-layer model in which a dual-path architectural model has been defined. For recognizing the gate in the first path using the proposed architectural model, the model is trained by the training dataset. But in the second path, the training dataset is first examined by a screening system and the best images containing features different from the features of the first path are selected. In another word, features of a network similar to the first path, but after passing through the screening system are generated in the second path. Afterwards, the feature vectors generated in two paths are combined to create a global feature vector, thus covering different spaces of the gate recognition problem. Different evaluations have been performed on the presented method. The evaluation results represent the improved accuracy of gate recognition using the proposed dual-path architectural model in comparison to the basic model. A comparison of proposed method with other existing outcomes also represents the improved accuracy of the proposed method in comparison to the published results.
P. Shi and X. Yu, “Goal event detection in soccer videos using multi-clues detection rules,” in Management and Service Science, 2009. MASS’09. International Conference on, 2009, pp. 1–4.
M.-L. Shyu, Z. Xie, M. Chen, and S.-C. Chen, “Video semantic event/concept detection using a subspace-based multimedia data mining framework,” IEEE Transactions on Multimedia, vol. 10, no. 2, pp. 252–259, 2008.
M. H. Kolekar, “Bayesian belief network based broadcast sports video indexing,” Multimedia Tools and Applications, vol. 54, no. 1, pp. 27–54, 2011.
D. W. Tjondronegoro and Y.-P. P. Chen, “Knowledge-discounted event detection in sports video,” Ieee transactions on systems, man, and cybernetics-part a: Systems and humans, vol. 40, no. 5, pp. 1009–1024, 2010.
L.-Y. Duan, M. Xu, Q. Tian, C.-S. Xu, and J. S. Jin, “A unified framework for semantic shot classification in sports video,” IEEE Transactions on Multimedia, vol. 7, no. 6, pp. 1066–1083, 2005.
B. Li, J. H. Errico, H. Pan, and I. Sezan, “Bridging the semantic gap in sports video retrieval and summarization,” Journal of Visual Communication and Image Representation, vol. 15, no. 3, pp. 393–424, 2004.
