When there exist not enough historical defect data for building an accurate prediction model, semisupervised defect prediction (SSDP) and cross-project defect prediction (CPDP) are two feasible solutions. Existing CPDP methods assume that the available source data are well labeled. However, due to expensive human efforts for labeling a large amount of defect data, usually, we can only utilize the suitable unlabeled source data. We call CPDP in this scenario as cross-project semisupervised defect prediction (CSDP). Although some within-project semisupervised defect prediction (WSDP) methods have been developed in recent years, there still exists much room for improvement on prediction performance. In this paper, we aim to provide a unified and effective solution for both CSDP and WSDP problems. We introduce the semisupervised dictionary learning technique and propose a cost-sensitive kernelized semisupervised dictionary learning (CKSDL) approach. CKSDL can make full use of the limited labeled defect data and a large amount of unlabeled data in the kernel space. In addition, CKSDL considers the misclassification costs in the dictionary learning process. Extensive experiments on 16 projects indicate that CKSDL outperforms state-of-the-art WSDP methods, using unlabeled cross-project defect data can help improve the WSDP performance, and CKSDL generally obtains significantly better prediction performance than related SSDP methods in the CSDP scenario.
The modality and pose variance between RGB and infrared (IR) images are two key challenges for RGB-IR person re-identification. Existing methods mainly focus on leveraging pixel or feature alignment to handle the intra-class variations and cross-modality discrepancy. However, these methods are hard to keep semantic identity consistency between global and local representation, which the consistency is important for the cross-modality pedestrian re-identification task. In this work, we propose a novel cross-modality graph reasoning method (CGRNet) to globally model and reason over relations between modalities and context, and to keep semantic identity consistency between global and local representation. Specifically, we propose a local modality-similarity module to put the distribution of modality-specific features into a common subspace without losing identity information. Besides, we squeeze the input feature of RGB and IR images into a channel-wise global vector, and through graph reasoning, the identity relationship and modality relationship in each vector are inferred. Extensive experiments on two datasets demonstrate the superior performance of our approach over the existing state-of-the-art. The code is available at https://github.com/fegnyujian/CGRNet .
1 February 2022 in Computers and Electrical Engineering
Vehicle pose estimation with LIDAR plays a crucial role in autonomous driving systems. It serves as the fundamental basis for functions such as tracking, path planning, and decision-making. However, the majority of current vehicle pose estimation techniques struggle to produce satisfactory results when faced with incomplete observation measurements, such as L-shaped point cloud clusters without side contours or those including side-view mirrors. In addition, the requirement for real-time results further increases the difficulty of the pose estimation task. In this paper, we present a vehicle Pose Estimation method with Heuristic L-shape fitting and grid-based Particle Filter (PE-HL-PF). We design a geometric shape classifier module to divide clusters into symmetrical and asymmetrical ones according to their shape features. Furthermore, a contour-based heuristic L-shape fitting module is introduced for asymmetrical clusters, and a structure-aware grid-based particle filter is used to estimate the pose of symmetrical clusters. PE-HL-PF first utilizes a heuristic asymmetrical module that selects dominant contours fitting orientation in a heuristic manner, thereby avoiding the need for a complex traversal search. Additionally, a symmetrical module based on particle filtering is incorporated to enhance the stability of orientation estimation. This method achieves significant improvements in both the runtime efficiency and pose estimation accuracy of incomplete point clouds. Compared with state-of-the-art pose estimation methods, our PE-HL-PF demonstrates a notable performance improvement. Our method can estimate the pose of thousands of objects in less than 1 millisecond, a significant improvement over previous methods. The results of experiments performed on the KITTI dataset validate the effectiveness of our approach.
Much research on adversarial attacks has proved that deep neural networks have certain security vulnerabilities. Among potential attacks, black-box adversarial attacks are considered the most realistic based on the the natural hidden nature of deep neural networks. Such attacks have become a critical academic emphasis in the current security field. However, current black-box attack methods still have shortcomings, resulting in incomplete utilization of query information. Our research, based on the newly proposed Simulator Attack, proves the correctness and usability of feature layer information in a simulator model obtained by meta-learning for the first time. Then, we propose an optimized Simulator Attack+ based on this discovery. Our optimization methods used in Simulator Attack+ include: (1) a feature attentional boosting module that uses the feature layer information of the simulator to enhance the attack and accelerate the generation of adversarial examples; (2) a linear self-adaptive simulator-predict interval mechanism that allows the simulator model to be fully fine-tuned in the early stage of the attack and dynamically adjusts the interval for querying the black-box model; and (3) an unsupervised clustering module to provide a warm-start for targeted attacks. Results from experiments on the CIFAR-10 and CIFAR-100 datasets clearly show that Simulator Attack+ can further reduce the number of consuming queries to improve query efficiency while maintaining the attack.
28 June 2023 in IEEE Transactions on Industrial Informatics
Light-detection-and-ranging-based multiobject detection and tracking play fundamental roles in autonomous driving systems. Most existing detection and tracking methods inevitably require complex pairing permutations for object association across frames, making the framework slow. Moreover, the occlusion and viewpoint changes lead to missed and false detection. To solve the abovementioned issues, this article proposes a simple joint 3-D detection and tracking approach with three-step offset learning ( $\text{Offset}^{3}\text{Net}$ ). Specifically, $\text{Offset}^{3}\text{Net}$ incorporates three task-specific output subnetworks to learn three offsets: 1) center offset, 2) motion offset, and 3) association offset. The learning of abovementioned offsets eliminates the complex bipartite matching processing. Specifically, the center offset guides the model to generate precise detections, whereas the motion offset transforms the track from the previous frame to the current frame, and the association offset minimizes the distance between detection and motion-updated track of the same object. Then, a simple read-off operation is conducted for data association on a hybrid-time centerness map, which represents the detections and offset-updated tracks. In addition, we design a detection-feature-enhanced module that captures the temporal coherence of the object motion and appearance information, avoiding the missed and false detection. Experiments on nuScenes have demonstrated the effectiveness of our $\text{Offset}^{3}\text{Net}$ in terms of accuracy and speed compared with most 3-D detection and tracking methods.