With the growing demand for video content analysis, sports video activity recognition has wide application prospects and commercial value, such as computer-assisted highlight extraction, tactic statistics and strategi...
详细信息
With the growing demand for video content analysis, sports video activity recognition has wide application prospects and commercial value, such as computer-assisted highlight extraction, tactic statistics and strategic analysis. Volleyball group activity recognition focuses on understanding the action performed by a group of players in volleyball matches. However, due to the cluttered backgrounds and the complex relationships between individuals in volleyball video, group activity recognition for sports video has become a significant and challenging issue. Therefore, we propose a dual attention based on a spatial-temporal inference network for volleyball group activity recognition. First, a dual attention model composed of spatial attention and mixture channel attention is proposed to assign attention weight dynamically to each feature and concern on the interdependencies of group members. It can improve the capacity of the model to distinguish features representation with intra-class variation by obtaining rich contextual relationships. Next, to focus on individual spatial-temporal information, an individual spatial-temporal inference network (ISTIN) is designed to capture person-group interactions for emphasizing the variability of these information. Finally, these features are fed into a recurrent neural network to capture temporal dependencies and make the classification. Experimental results show that this approach can be effective in group activity recognition, with our model improving recognition rates over baseline method on the benchmark datasets: Volleyball dataset and Collective Activity dataset.
Forecasting the accessibility of urban mobility infrastructure, such as transport networks and charging places for vehicles, is crucial to our daily lives. However, the event records of the mobility infrastructure in ...
详细信息
ISBN:
(纸本)9789819755516;9789819755523
Forecasting the accessibility of urban mobility infrastructure, such as transport networks and charging places for vehicles, is crucial to our daily lives. However, the event records of the mobility infrastructure in a region can be unarchived due to low sensor coverage or different jurisdictions. This hinders accurate assessment and leads to a waste of resources. We target the problem of spatial-temporal inference of events with zero historical and real-time in-region records in urban mobility infrastructures. The difficulties lie in the limited knowledge of events in the target region and the lack of generalizability across different regions and distinct mobility infrastructures. To address the problem, we propose a hierarchy-driven machine-learning approach that exploits the hierarchical transport networks as domain-general features and heterogeneous urban sensory data as domain-specific features to support cross-region inference without directly using historical and real-time event records. We evaluate our approach on eight real-world datasets with two downstream tasks: charging availability and road accessibility. The evaluation results show the efficacy and adaptability of our approach by consistently achieving statistically significant improvement over state-of-the-art methods.
The vehicle instrument cluster is one of the most advanced and complicated electronic embedded control systems used in modern vehicles providing a driver with an interface to control and determine the status of the ve...
详细信息
The vehicle instrument cluster is one of the most advanced and complicated electronic embedded control systems used in modern vehicles providing a driver with an interface to control and determine the status of the vehicle. In this paper, we develop a novel hybrid approach called Hierarchical spatial-temporal State Machine (HSTSM). The approach addresses a problem of spatial-temporal inference in complex dynamic systems. It is based on a memory-prediction framework and Deep Neural Networks (DNN) which is used for fault detection and isolation in automatic inspection and manufacturing of vehicle instrument cluster. The technique has been compared with existing methods namely rule-based, template-based, Bayesian, restricted Boltzmann machine and hierarchical temporal memory methods. Results show that the proposed approach can successfully diagnose and locate multiple classes of faults under real-time working conditions.
Recently, the video retrieval processing is concerned with retrieving videos that are relevant to the users' requests from a large collection of videos, referred to as a video database. We have proposed 3D Z-strin...
详细信息
ISBN:
(纸本)9781424442843
Recently, the video retrieval processing is concerned with retrieving videos that are relevant to the users' requests from a large collection of videos, referred to as a video database. We have proposed 3D Z-string to represent symbolic videos accompanying with the string generation and video reconstruction algorithms. In this paper, we proposed the spatial-temporal similarity retrieval approach of vides in 3D Z-string. Our approach defines a set of user assigned weights, based on the factors of spatial-temporal relations of object pairs in a video, in order to rank the retrieval videos. We use dynamic programming to calculate the similarity measures and propose the similarity retrieval algorithm. By providing various criterion of similarity between videos to match user requirement, our proposed similarity retrieval algorithm has discrimination power about different criteria.
暂无评论