SkateFormer: Skeletal-Temporal Transformer for Human Action Recognition
Efficiently capturing skeletal-temporal relations for improved action recognition.
Leveraging Large Language Models for Accurate Skeleton-based Action Recognition
Large language models can be effectively leveraged as powerful action recognizers by projecting skeleton sequences into "action sentences" that are compatible with the models' pre-trained knowledge.
Spiking Graph Convolutional Network with Multimodal Fusion and Knowledge Distillation for Efficient Skeleton-based Action Recognition
A novel Spiking Graph Convolutional Network (SGN) with multimodal fusion and knowledge distillation is proposed to achieve efficient and accurate skeleton-based action recognition.
A Hybrid Dual-Branch Network for Robust Skeleton-Based Action Recognition
The proposed Hybrid Dual-Branch Network (HDBN) effectively combines Graph Convolutional Networks (GCNs) and Transformers to achieve robust and accurate skeleton-based action recognition.
Improved Graph Pooling Network for Efficient Skeleton-Based Action Recognition
The proposed Improved Graph Pooling Network (IGPN) incorporates a region-aware pooling strategy, cross fusion block, and information supplement module to enhance the representation ability of skeleton features while reducing computational overhead.
Enhancing Action Recognition from Low-Quality Skeleton Data via Part-Level Knowledge Distillation
The proposed knowledge distillation framework distills discriminative part-level knowledge from heterogeneous high-quality skeletons to enhance representations of low-quality skeletons, enabling accurate action recognition even in the presence of intensive noise.
Efficient Skeleton-Based Action Recognition through Convolutional Transformers and Self-Supervised Learning
ReL-SAR, a lightweight convolutional transformer model, leverages self-supervised learning with BYOL to extract robust and generalizable features from skeleton sequences for efficient action recognition.
© 2024 by Linnk AI