No abstract

CNN is particularly effective in extracting spatial features. However, the single-layer classifier constructed by activation function in CNN is easily interfered by image noise, resulting in reduced classification accuracy. To solve the problem, the advanced ensemble model XGBoost is used to overcome the deficiency of a single classifier to classify image features. To further distinguish the extracted image features, a CNN-XGBoost image classification model optimized by APSO is proposed, where APSO optimizes the hyper-parameters on the overall architecture to promote the fusion of the two-stage model. The model is mainly composed of two parts: feature extractor CNN, which is used to automatically extract spatial features from images; feature classifier XGBoost is applied to classify features extracted after convolution. In the process of parameter optimization, to overcome the shortcoming that traditional PSO algorithm easily falls into a local optimal, the improved APSO guide the particles to search for optimization in space by two different strategies, which improves the diversity of particle population and prevents the algorithm from becoming trapped in local optima. The results on the image set show that the proposed model gets better results in image classification. Moreover, the APSO-XGBoost model performs well on the credit data, which indicates that the model has a good ability of credit scoring.

No abstract

No abstract

Human action recognition has become a challenging task in computer vision because it is difficult to combine spatiotemporal information. A multi-attention spatiotemporal graph convolution network is proposed.The core idea is to construct a connected graph according to the time series information and natural connection of human skeleton, and use the spatiotemporal graph convolution network with multi-attention mechanism to automatically learn spatial and temporal features and optimize the connected graph to realize prediction. Graph attention module is introduced, the topological structure of the graph constructed by the model will be optimized with the process of network training after initialization, then the topological structure which is more suitable for expressing human actions will be obtained. In addition, the channel attention module is added to make the network pay more attention to the important channel information, so as to extract the features of describing actions more effectively. A large number of experiments are carried out on the recognized large datasets: NTU-RGDB and Kinectics, which show that the method has higher recognition accuracy.