Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-10 Cooperative journals: 《计算机应用研究》
Abstract： Action recognition(AR) is a hot research area in computer vision field, and has an extensive application prospect for security monitoring, autopilot, production safety etc. Firstly, this paper analysed the connotation and denotation of AR and put forward the technical challenges; Secondly, the paper analysed and compared the working principles of AR from three aspects: time feature extraction, efficient optimization and long-term feature capture; Thirdly, in order to select suitable AR models for different application scenarios, this paper compared the performance characterization of 43 benchmark AR methods in recent ten years based on UCF101, HMDB51, Something-Something and Kinetics400 data sets. Finally, this paper pointed out the future development direction of AR field, and the research results can provide theoretical reference and technical support for video feature extraction and visual content understanding.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-10 Cooperative journals: 《计算机应用研究》
Abstract： In recent years, the deep supervised hash retrieval method have been successfully applied to many image retrieval systems. However, the existing methods still have some shortcomings: First, most of the deep hash learning methods use symmetric strategies to train the network, but the training of this strategy is usually time-consuming and difficult to be used in the large-scale hash learning process; Second, there is a discrete optimization problem in the hash learning process. Existing methods relax this problem and it is difficult to guarantee the optimal solution. In order to solve the above problems, this paper proposes a greedy-asymmetric deep supervised hashing method for image retrieval, which fully combines the advantages of the greedy algorithm and asymmetric strategy to further improve the hash retrieval performance. This article compares 17 state-of-the-art methods on two commonly used datasets. Compared with the state-of-the-art methods, our method increases the mAP in 48-bits setting by 1.3% on CIFAR-10 dataset. And on NUS-WIDE dataset, increases the mAP in all-bits setting by increased 2.3% on average. The experimental results show that our method can further improve the performance of hash retrieval.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： In order to improve the accuracy of truth discovery in multi-truth scene, this paper proposed a multi-ant colonies synchronization optimization based multi-truth discovery (MAC-SO-MTD) algorithm. It modeled the multi-truth discovery problem as the subset problem, which goal was maximizing the weighted sum of similarity between the set of observations provided by each data source and the set of true values of the object. On this basis, then designed ant colony algorithm to solve the problem. It set ant colonies according to the number of objects. Based on the subset problem’s structure graph, this paper used routes’ probability transition equations to search for truths synchronically. After one cycle, the best route of this cycle updating and no updating were two instances of updating pheromone, which improved the convergence speed. Finally, the analysis of algorithm complexity and contrast experiment on the real data set validated the superiority of the algorithm.
Hits
Hits
Downloads
Comment