6 Hits in 0.9 sec

Skeleton Aware Multi-modal Sign Language Recognition [article]

Songyao Jiang, Bin Sun, Lichen Wang, Yue Bai, Kunpeng Li, Yun Fu
2021 arXiv   pre-print
Recently, skeleton-based action recognition attracts increasing attention due to the independence between the subject and background variation.  ...  RGB and depth modalities are also incorporated and assembled into our framework to provide global information that is complementary to the skeleton-based methods SL-GCN and SSTCN.  ...  The decoupling GCN module, the dropgraph module and the STC attention mechanism all contribute to our final recognition rate. Table 4 .  ... 
arXiv:2103.08833v5 fatcat:dl7aebwtxbam5ftpjkgaaa5pae

Constructing Stronger and Faster Baselines for Skeleton-based Action Recognition [article]

Yi-Fan Song, Zhang Zhang, Caifeng Shan, Liang Wang
2022 arXiv   pre-print
) baseline for skeleton-based action recognition.  ...  One essential problem in skeleton-based action recognition is how to extract discriminative features over all skeleton joints.  ...  Compared to the first GCN baseline for skeleton-based action recognition, i.e., ST-GCN [11] , EfficientGCN-B0 outperforms by 8.7% in accuracy, with a 5.98× fewer FLOPs and a 10.68× fewer parameters.  ... 
arXiv:2106.15125v2 fatcat:4d623jwkx5bqfmwgivc25rv3jq

Topology-aware Convolutional Neural Network for Efficient Skeleton-based Action Recognition [article]

Kailin Xu, Fanfan Ye, Qiaoyong Zhong, Di Xie
2021 arXiv   pre-print
In the context of skeleton-based action recognition, graph convolutional networks (GCNs) have been rapidly developed, whereas convolutional neural networks (CNNs) have received less attention.  ...  Compared with leading GCN-based methods, we achieve comparable performance with much less complexity in terms of the required GFLOPs and parameters.  ...  Decoupling GCN with DropGraph Module for Skeleton-Based Ac- tion Recognition. In Vedaldi, A.; Bischof, H.; Brox, T.; and Frahm, Liu, M.; Liu, H.; and Chen, C. 2017. Enhanced skeleton visualiza- J.  ... 
arXiv:2112.04178v2 fatcat:zwbupfy76nd7dnjqeaotii66ye

Tripool: Graph triplet pooling for 3D skeleton-based action recognition

Wei Peng, Xiaopeng Hong, Guoying Zhao
2021 Pattern Recognition  
Our architecture is also based on this module. Nonetheless, in skeleton-based action recognition, graph pooling operations are rarely found in current state-of-the-art GCN architectures.  ...  Graph Convolutional Network (GCN) has already been successfully applied to skeleton-based action recognition.  ...  To efficiently model the spatial-temporal skeleton and avoid overfitting problem, Cheng et al. proposed to decouple GCN to boost the graph modeling ability and introduce DropGraph module to discard features  ... 
doi:10.1016/j.patcog.2021.107921 fatcat:75paioovd5dmpgiigxmwfggo7e

OpenHands: Making Sign Language Recognition Accessible with Pose-based Pretrained Models across Languages [article]

Prem Selvaraj, Gokul NC, Pratyush Kumar, Mitesh Khapra
2021 arXiv   pre-print
Second, we train and release checkpoints of 4 pose-based isolated sign language recognition models across all 6 languages, providing baselines and ready checkpoints for deployment.  ...  We introduce OpenHands, a library where we take four key ideas from the NLP community for low-resource languages and apply them to sign languages for word-level recognition.  ...  Our extended gratitude also goes to Zenodo, who helped us with hosting our large datasets (NC and Selvaraj 2021).  ... 
arXiv:2110.05877v1 fatcat:3zawbhrfsbgdnczpmzzb3jqy4a

Sign Language Recognition via Skeleton-Aware Multi-Model Ensemble [article]

Songyao Jiang, Bin Sun, Lichen Wang, Yue Bai, Kunpeng Li, Yun Fu
Recently, skeleton-based action recognition has attracted increasing attention due to its subject-invariant and background-invariant nature, whereas skeleton-based SLR is still under exploration due to  ...  The skeleton-based predictions are fused with other RGB and depth based modalities by the proposed late-fusion GEM to provide global information and make a faithful SLR prediction.  ...  Besides, we find that the DropGraph module, the decoupling GCN module, and the STC attention mechanism all contribute to the final performance. D.  ... 
doi:10.48550/arxiv.2110.06161 fatcat:mfq7t3peurg33di6i35wrgzqtu