Action recognition via pose-based graph convolutional networks with intermediate dense supervision

Shi, Lei; Zhang, Yifan; Cheng, Jian; Lu, Hanqing

doi:10.1016/j.patcog.2021.108170

Cited by 26 publications

(5 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Graph Convolutional Network (GCN) has been used in several computer vision tasks such as action recognition [33], brain disorder prediction [34], image retrieval [35], person re-identification [36], and recommendation systems [37]. There are different types of GCN which have been introduced based on various applications.…”

Section: Graph Convolutional Networkmentioning

confidence: 99%

See 1 more Smart Citation

TAA-GCN: A temporally aware Adaptive Graph Convolutional Network for age estimation

Korban

Youngs²,

Acton³

2023

Pattern Recognition

View full text Add to dashboard Cite

Section: Graph Convolutional Networkmentioning

confidence: 99%

“…For different tasks, the researchers designed various GCN architectures. For example, [33] designed a human pose-aware GCN to model the dependencies among human skeleton joints and body parts. [34] proposed a Hierarchical GCN to learn from different ROIs in fMRI data of the brain.…”

Section: Graph Convolutional Networkmentioning

confidence: 99%

TAA-GCN: A temporally aware Adaptive Graph Convolutional Network for age estimation

Korban

Youngs²,

Acton³

2023

Pattern Recognition

View full text Add to dashboard Cite

“…They proposed a convolution-free spatio-temporal Transformer network to extract features. Lei et al [29] represented a posebased graph convolutional networks, which use both video data and skeleton data to model.…”

Section: A Skeleton-based Action Recognitionmentioning

confidence: 99%

MSA-GCN:Multiscale Adaptive Graph Convolution Network for Gait Emotion Recognition

Yin¹,

Li²,

Huang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Gait emotion recognition plays a crucial role in the intelligent system. Most of the existing methods recognize emotions by focusing on local actions over time. However, they ignore that the effective distances of different emotions in the time domain are different, and the local actions during walking are quite similar. Thus, emotions should be represented by global states instead of indirect local actions. To address these issues, a novel MultiScale Adaptive Graph Convolution Network (MSA-GCN) is presented in this work through constructing dynamic temporal receptive fields and designing multiscale information aggregation to recognize emotions. In our model, a adaptive selective spatial-temporal graph convolution is designed to select the convolution kernel dynamically to obtain the soft spatiotemporal features of different emotions. Moreover, a Cross-Scale mapping Fusion Mechanism (CSFM) is designed to construct an adaptive adjacency matrix to enhance information interaction and reduce redundancy. Compared with previous state-of-the-art methods, the proposed method achieves the best performance on two public datasets, improving the mAP by 2%. We also conduct extensive ablations studies to show the effectiveness of different components in our methods.

show abstract

“…2D human pose estimation (HPE) based on input images is a significant and challenging hotspot in computer vision. It has many applications, such as action recognition, 1 tracking, 2 and fitness, 3 and it is also the basis of video pose estimation 4 and 3D HPE 5 . With the rapid development of mobile devices, the demand for HPE networks with small model sizes, low complexity, and high accuracy has been increasing.…”

Section: Introductionmentioning

confidence: 99%

FALNet: flow-based attention lightweight network for human pose estimation

Xiao,

Liu,

2023

J. Electron. Imag.

View full text Add to dashboard Cite

Deep neural networks achieve significant progress in human pose estimation (HPE), but there are still many difficulties in practical applications. One of the important reasons is that the current advanced models are very complex leading to huge computation cost. We propose a flow-based attention lightweight network (FALNet) for HPE to solve this problem. We first design a cheap lightweight bottleneck to reduce the model size and complexity with two components: depthwise convolution and cheap unit. Then we propose a flow-based fusion attention block to generate and aggregate multi-scale features of the same layer effectively. We demonstrate the effectiveness of our methods on two benchmark datasets: the COCO dataset and the MPII dataset. Our network FALNet-50 only has 2.2 M parameters and 0.66G floating-point operations (FLOPs), achieving comparable or even better accuracy with smaller complexity. Moreover, we show the speed advantage of our network during inference. Specifically, our FALNet-50 achieves 70.4 in average precision score on COCO val2017 with 16 FPS inference speed on a smartphone.

show abstract

Action recognition via pose-based graph convolutional networks with intermediate dense supervision

Cited by 26 publications

References 3 publications

TAA-GCN: A temporally aware Adaptive Graph Convolutional Network for age estimation

TAA-GCN: A temporally aware Adaptive Graph Convolutional Network for age estimation

MSA-GCN:Multiscale Adaptive Graph Convolution Network for Gait Emotion Recognition

FALNet: flow-based attention lightweight network for human pose estimation

Contact Info

Product

Resources

About