Search

Graph convolutional networks (GCNs) have attracted considerable interest in skeleton-based action recognition. Existing GCN-based models have proposed methods to learn dynamic graph topologies generated from the feature information of vertices to capture inherent relationships. However, these models have two main limitations. Firstly, they struggle to effectively utilize high-dimensional or structural information, which limits their capacity for feature representation and consequently hinders performance improvement. Secondly, among these models, the multi-scale methods that aggregate information at different scales often over-capture unnecessary relationships between vertices. This leads to an over-smoothing problem where smoothed features are extracted, making it difficult to distinguish the features of each vertex. To address these limitations, we propose the multi-scale structural graph convolutional network (MSS-GCN) for skeleton-based action recognition. Within the MSS-GCN framework, the common intersection graph convolution (CI-GC) leverages the overlapped neighbor information, indicating the overlap between neighboring vertices for a given pair of root vertices. The graph topology of CI-GC is designed to compute the structural correlation between neighboring vertices corresponding to each hop, thereby enriching the context of inter-vertex relationships. Then, our proposed multi-scale spatio-temporal modeling aggregates local-global features to provide a comprehensive representation. In addition, we propose a Graph Weight Annealing (GWA) method, which is a graph scheduling method to mitigate the over-smoothing caused by multi-scale aggregation. By varying the importance between a vertex and its neighbors, we demonstrate that the over-smoothing problem can be effectively mitigated. Moreover, our proposed GWA method can easily be adapted to different GCN models to enhance performance. Combining the MSS-GCN model and the GWA method, we propose a powerful feature extractor that effectively classifies actions for skeleton-based action recognition in various datasets. We evaluate our approach on three benchmark datasets: NTU RGB+D, NTU RGB+D 120, and NW-UCLA. The proposed MSS-GCN achieves state-of-the-art performance on all three datasets, further validating the effectiveness of our approach.

We describe a face recognition system based on two different three-dimensional (3D) sensors. We use 3D sensors to overcome the pose-variation problems that cannot be effectively solved in two-dimensional images. We acquire input data based on a structured-light system and compare it with 3D faces that are obtained from a 3D laser scanner. Owing to differences in structure between the input data and the 3D faces, we can generate the range images of the probe and stored images. For estimating the head pose of input data, we propose a novel error-compensated singular-value decomposition that geometrically estimates the rotation angle. Face recognition rates obtained with principal component analysis on various range images of 35 people in different poses show promising results.

We propose a preprocessing system that transforms the real world into 3D mesh virtual world to provide augmented reality (AR) services. The proposed system uses a monocular RGB camera to obtain sequential images to generate real space. First, we create 3D real-world space composed of point clouds using sequential color images. And then, we segment the objects in each color image and assign an identification number for each object. Also, we execute ‘3D point labeling’, which assigns identification numbers from 2D segmented object to 3D point cloud through a simple projection manner with several parameters. This process could collect 3D point cloud with same the label by as an object. Finally, only the 3D point clouds that have the same identification number by as an object are used to generate 3D mesh. This paper confirmed that 3D mesh could be created using the only general monocular camera without the help of special 3D sensors.