Search

Only parts of unlabeled data are selected to train models for most semi-supervised learning methods, whose confidence scores are usually higher than the pre-defined threshold (i.e., the confidence margin). We argue that the recognition performance should be further improved by making full use of all unlabeled data. In this paper, we learn an Adaptive Confidence Margin (Ada-CM) to fully leverage all unlabeled data for semi-supervised deep facial expression recognition. All unlabeled samples are partitioned into two subsets by comparing their confidence scores with the adaptively learned confidence margin at each training epoch: (1) subset I including samples whose confidence scores are no lower than the margin; (2) subset II including samples whose confidence scores are lower than the margin. For samples in subset I, we constrain their predictions to match pseudo labels. Meanwhile, samples in subset II participate in the feature-level contrastive objective to learn effective facial expression features. We extensively evaluate Ada-CM on four challenging datasets, showing that our method achieves state-of-the-art performance, especially surpassing fully-supervised baselines in a semi-supervised manner. Ablation study further proves the effectiveness of our method. The source code is available at https://github.com/hangyu94/Ada-CM., Accepted by CVPR2022

Obtaining the accurate position and attitude of the object is the key to realize rendezvous and docking, on-orbit maintenance and other related tasks of space spacecraft. However, for non-cooperative objects, they lack prearranged cooperation sign, which would make it much more difficult to estimate their pose. Therefore, this paper attempts to use the powerful feature learning ability of neural network to establish the mapping relationship between the object in image and its current pose, then regresses the pose parameters of the object from a monocular image. Finally, we tested and verified the network on the public satellite dataset called Speed. The results showed that the translation error was 0.1237 and the rotation error was 0.1335.

Clustering by fast search and find of density peaks published on journal Science in 2014 is a density-based clustering technique, which is not only unnecessary to determine the number of clusters in advance, but also able to recognize the clusters of arbitrary shapes. Due to a manual selection of clustering centers on a decision graph, samples which belong to one cluster may be assigned to two or more clusters and vice versa. On assumption that boundary points which keep comparable densities with cluster centers should be regarded as inner points, we make a new method which not only can find all possible clusters automatically but also can combine those with similarities simultaneously to obtain the final clusters. Unlike clustering by fast search and find of density peaks, we only focus on densities with discarding the relative metric which measures the minimum distance between a cluster center and a point with a higher density. Qualitative and quantitative experimental results on sufficient datasets demonstrate the effectiveness of our method.

The majority of effective information on express profile data remains under-utilized due to the characteristics of high-dimensional features, complex correlations among sample attributes and the lack of skills and resources to analyze and visualize these data. In this paper, we developed an automatic feature selection platform combined with an intuitive and interactive interface for differential expression analysis (abbreviated as AFSDEA), which utilized ensemble learning to automatically select features on expression profile data. Qualitative and quantitative analyses indicating the interpretable and predictive ability of the selected features were provided. Correspondingly, experimental results on simulated and real data proved the effectiveness of the developed platform. The establishment of this platform generates a custom Web browser for automatically selecting features and provides outputs for differential expression analysis, which will facilitate the research on expression profile data. The webserver of AFS-DEA is available at http://bio-nefu.com/afs-dea.

In this paper, the differential equation model and difference equation model of furnace temperature curve based on heat radiation principle are established by using the knowledge of thermodynamics. The least square method is used to process the experimental data of furnace temperature curve, and the unknown parameters in the model is obtained. Then the theoretical value of the furnace temperature curve is calculated by using the determined difference equation model. The results show that the theoretical values are in good agreement with the experimental values, which verifies the rationality of the established model. Finally, the optimal selection method of maximum furnace speed based on dichotomy is proposed in combination with the model.

Runway incursion has always been a key factor affecting the safety of runway operation. Most high-hazard runway conflicts are time-critical, in which the best mitigation strategy is a direct warning to the pilot that the runway ahead is unsafe for crossing or for takeoff. Runway Status Light System is the most effective runway incursion prevention system which using red light to give the pilot a direct warning of runway status. In this paper, we propose a runway status light control strategy and implementation method based on state transition. Firstly, we discuss the components of the runway status light system and the basic principles of its control logical. Then, we establish a state transition model for aircraft and vehicles, and define the control method for each target state of the model in combination with airport ground operation rules. Finally, we validated that the proposed control model can meet the operation requirements using the demonstration results at Hongqiao Airport.

An accurate and robust altitude controller is critical for vertical takeoff and landing (VTOL) unmanned aerial vehicles (UAVs) in achieving quasi-stationary flight. Most UAV altitude control designs neglect the rotor dynamics. Therefore, they cannot be used for a tail-sitter UAV equipped with turbine engines because of the complicated engine dynamics with an apparent time delay. In this paper, we develop an integrated altitude controller that considers the engine dynamics. The new controller consists of a proportional-derivative (PD) control term and an acceleration feedback term. The stability region in the parameter space is analyzed and the controller is designed to achieve specific gain and phase margins. A UAV hover flight experiment is conducted and the results are presented to demonstrate the effectiveness of the proposed altitude controller.

Similarity join of two datasets P and Q is a primitive operation that is useful in many application domains. The operation involves identifying pairs (p, q), in the Cartesian product of P and Q such that (p, q) satisfies a stipulated similarity condition. In a high-dimensional space, an approximate similarity join based on locality-sensitive hashing (LSH) provides a good solution while reducing the processing cost with a predictable loss of accuracy. A distributed processing framework such as MapReduce allows the handling of large and high-dimensional datasets. However, network cost frequently turns into a bottleneck in a distributed processing environment, thus resulting in a challenge of achieving faster and more efficient similarity join [2]. This paper focuses on collision counting LSH-based similarity join in MapReduce and proposes a network-efficient solution called C2Net to improve the utilization of MapReduce combiners. The solution uses two graph partitioning schemes: (i) minimum spanning tree for organizing LSH buckets replication; and (ii) spectral clustering for runtime collision counting task scheduling. Experiments have shown that, in comparison to the state of the art, the proposed solution is able to achieve 20% data reduction and 50% reduction in shuffle time.

Grinding circuit should provide stable particle size distribution while operating in a way to maximize mill efficiency. In this paper, a fuzzy logic-based supervisory control strategy is proposed for controlling product particle size while improving mill efficiency in a ball mill grinding circuit. In the supervisory level, fuzzy logic-based control determines the optimized setpoints of the regulatory level controllers. The whole system ensures a long-term stableness of product quality while increasing the mill efficiency more than six percent in practical industrial application in comparison with conventional control. The results of the practical industrial operation demonstrate the feasibility, reliability and effectiveness of the proposed control strategy.