GCANet: Geometry cues-aware facial expression recognition based on graph convolutional networks

Wang, Shutong; Zhao, Anran; Lai, Chenghang; Zhang, Qi; Li, Duantengchuan; Gao, Yihua; Dong, Liangshan; Wang, Xiaoguang

doi:10.1016/j.jksuci.2023.101605

Cited by 4 publications

(1 citation statement)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Because of the success of GNN methods in recommender systems and social networks [23][24][25][26], related research has been gradually conducted in computer vision [27][28][29][30][31][32]. In image annotation, Curve-GCN [28] used GCN for the fine prediction of labeled object contours, which improved the efficiency of image annotation; in image multi-label prediction, Chen et al [29] uncovered the number of times a combination of labels appeared in an image and combined with GCN to construct a link between multiple labels in an image, which assisted in the prediction of multiple labels; in facial expression recognition; GCANet [32] statistically analyzed the dataset and constructed a graph between AUs, and it used GCN to obtain the relationship between the composition of AUs and the corresponding emotions, which improved the accuracy of expression recognition. To prove that GNNs alone are also effective in processing images, Vision GNN [31] imitated Vision Transformer to perform the segmentation of images and constructed a graph network between these slices to selectively perform feature fusion between the slices, which is more flexible than the traditional method.…”

Section: Graph Neural Networkmentioning

confidence: 99%

DSPose: Dual-Space-Driven Keypoint Topology Modeling for Human Pose Estimation

Zhao,

Li,

Zeng

et al. 2023

Sensors

View full text Add to dashboard Cite

Human pose estimation is the basis of many downstream tasks, such as motor intervention, behavior understanding, and human–computer interaction. The existing human pose estimation methods rely too much on the similarity of keypoints at the image feature level, which is vulnerable to three problems: object occlusion, keypoints ghost, and neighbor pose interference. We propose a dual-space-driven topology model for the human pose estimation task. Firstly, the model extracts relatively accurate keypoints features through a Transformer-based feature extraction method. Then, the correlation of keypoints in the physical space is introduced to alleviate the error localization problem caused by excessive dependence on the feature-level representation of the model. Finally, through the graph convolutional neural network, the spatial correlation of keypoints and the feature correlation are effectively fused to obtain more accurate human pose estimation results. The experimental results on real datasets also further verify the effectiveness of our proposed model.

show abstract