Multilabel data share important features, including label imbalance, which has a significant influence on the performance of classifiers. Because of this problem, a widely used multilabel classification algorithm, the multilabel k-nearest neighbor (ML-kNN) algorithm, has poor performance on imbalanced multilabel data. To address this problem, this study proposes an improved ML-kNN algorithm based on value and weight. In this improved algorithm, labels are divided into minority and majority, and different strategies are adopted for different labels. By considering the label of latent information carried by the nearest neighbors, a value calculation method is proposed and used to directly classify majority labels. Additionally, to address the misclassification problem caused by a lack of nearest neighbor information for minority labels, weight calculation is proposed. The proposed weight calculation converts distance information with and without label sets in the nearest neighbors into weights. The experimental results on multilabel datasets from different benchmarks demonstrate the performance of the algorithm, especially for datasets with high imbalance. Different evaluation metrics show that the results are improved by approximately 2–10%. The verified algorithm could be applied to a multilabel classification of various fields involving label imbalance, such as drug molecule identification, building identification, and text categorization.

Label imbalance is one of the characteristics of multilabel data, and imbalanced data seriously affects the performance of the classifiers. In multilabel classification, resampling methods are mostly used to deal with imbalanced problems. Existing resampling methods balance the data by either undersampling or oversampling, which causes overfitting and information loss. Resampling has a significant impact on the minority labels. Furthermore, the high concurrency of majority labels and minority labels in many instances also affects the performance of classification. In this study, we proposed a bidirectional resampling method to decouple multilabel datasets. On one hand, the concurrency of labels can be reduced by setting termination conditions for decoupling, and on the other hand, the loss of instance information and overfitting can be alleviated by combining oversampling and undersampling. By measuring the minority labels of the instances, the instances that have less impact on minority labels are selected to resample. The number of resampling is limited to keep the original distribution of the data during the resampling phase. The experiments on seven benchmark multilabel datasets have proved the effectiveness of the algorithm, especially on datasets with high concurrency of majority labels and minority labels.

Attention is an important field to explore the importance of each convolutional kernel channel/weight. The existing attention methods mostly use the Squeeze-and-Excitation (SE) technology to extract the global nonlinear feature vectors as the weights of corresponding feature maps. However, the pooling operators and fully-connected layers used in SE technology extract global features at the cost of much valuable information loss and the parameter amount increase. Actually, the feature map containing full information is a ready-made and better attention for other feature maps in the same layer. Simultaneously the products of feature maps will bring powerful non-linearity. Seeing this, Kernel Product (KP) technology is proposed to simply get useful nonlinear attention. To verify the effectiveness of KP, the proposed KP module is employed on Selective Kernel Networks (SKNets) to take the place of the original SE technology. The variety of SKNets is called Kernel Product Networks (KPNets) in this paper. In addition, identity mapping is used to solve the non-convergence problem in very deep neural networks. The KPNets are evaluated on ImageNet-1k, CIFAR-10, and CIFAR-100. The experiment results show that KPNets outperform many state-of-the-art methods and get a similar but more efficient performance than its SKNets with counterpart.

Graph Pooling technology plays an important role in graph node classification tasks. Sorting pooling technologies maintain large-value units for pooling graphs of varying sizes. However, by analyzing the statistical characteristic of activated units after pooling, we found that a large number of units dropped by sorting pooling are negative-value units that contain representative information and can contribute considerably to the final decision. To maintain more representative information, we proposed a novel pooling technology, called Geometric Pooling (GP), containing the unique node features with negative values by measuring the similarity of all node features. We reveal the effectiveness of GP from the entropy reduction view. The experiments were conducted on TUdatasets to show the effectiveness of GP. The results showed that the proposed GP outperforms the SOTA graph pooling technologies by $1\%~\sim ~5\%$ with fewer parameters.

Multi-personpose estimation is a fundamental technology for deep learning-based intelligent sensing systems, enabling downstream understanding of human action in applications such as surveillance, robotics, healthcare, and sports analytics. Most two-stage multi-person pose estimation algorithms suffer from low efficiency due to their decoupled representation between body and keypoints, resulting in a complex inference process. Single-stage algorithms such as AdaptivePose that represent body parts as semantic proxy points can substantially simplify the pipeline while achieving competitive results; however, the limited receptive field of convolutional features still makes it difficult to localize proxy points for long-range extremities in semantically informative regions. IHPP addresses this issue as a task-specific refinement of AdaptivePose; it enriches the proxy-point perceiver with directional context modeling along the horizontal and vertical axes and redesigns the second-step regression stage with a part-wise branch to reduce inter-part feature interference while keeping the overall pipeline lightweight. Here, we present IHPP in three sizes (IHPP-S, IHPP-M, and IHPP-L), optimizing the tradeoffs between efficiency and accuracy through fine-tuning of the channel dimensions for each bodypart feature. IHPP-L achieves 72.3 AP at 28 fps on COCO test-dev, surpassing DEKR-W48 and SWAHR-W48 by 1.3 AP and 0.3 AP, respectively, while IHPP-M reaches 69.0 AP at 35 fps and IHPP-S runs at 42 fps with only 9.4 M parameters. On CrowdPose, IHPP-M outperforms AdaptivePose-W48 by 0.2 AP with about one-third the parameters. Comprehensive experiments on the MS COCO and CrowdPose datasets validate the effectiveness of this design.