가상 연결을 활용한 골격 기반 인간 행동 인식을 위한 적응형 하이퍼그래프 convolutional 네트워크
본 논문에서는 골격 데이터에서 여러 관절 간의 복잡한 상호 작용을 효과적으로 포착하는 하이퍼그래프를 활용하여 인간 행동 인식을 위한 새로운 접근 방식을 제안합니다.
Hyper-Graph Convolutional Networks for Skeleton-Based Action Recognition: Leveraging Virtual Connections and Adaptive Topology
This paper introduces Hyper-GCN, a novel method for skeleton-based action recognition that leverages hyper-graphs with virtual connections to capture complex multi-joint relationships and enhance feature aggregation for improved performance.
Extended Multi-Stream Temporal-Attention Adaptive Graph Convolutional Network (EMS-TAGCN) for Improved Skeleton-Based Human Action Recognition
This research proposes a novel graph convolutional network (GCN) architecture called EMS-TAGCN (Extended Multi-stream Temporal-attention Adaptive GCN) for skeleton-based human action recognition (HAR) that outperforms previous methods by incorporating bone information, adaptive graph topology, and a spatial-temporal-channel attention mechanism.
자동 회귀 적응형 하이퍼그래프 트랜스포머를 활용한 골격 기반 행동 인식
본 논문에서는 자동 회귀 적응형 하이퍼그래프 트랜스포머(AutoregAd-HGformer) 모델을 제안하여 골격 기반 행동 인식에서 기존 방법보다 뛰어난 성능을 달성했습니다. 이는 모델 내부 및 외부에서 하이퍼그래프를 생성하는 두 가지 새로운 기술과 다양한 attention 메커니즘을 통해 이루어졌습니다.
Autoregressive Adaptive Hypergraph Transformer (AutoregAd-HGformer) for Improved Skeleton-based Action Recognition
This research paper introduces AutoregAd-HGformer, a novel hypergraph transformer architecture that leverages adaptive hypergraph generation and multi-level attention mechanisms to achieve state-of-the-art performance in skeleton-based action recognition.
Human Action Recognition Using a Novel Skeleton-based Quantum Spatial Temporal Relative Transformer Network (ST-RTR)
This research proposes a novel Spatial-Temporal Relative Transformer Network (ST-RTR) for skeleton-based human action recognition, leveraging quantum-inspired computing principles to enhance performance and overcome limitations of existing methods like ST-GCN.
서로 다른 데이터셋 간의 골격 행동 인식을 위한 완전한 행동 복구 및 재샘플링 증강 기법
서로 다른 데이터셋에서 수집된 골격 행동 데이터 간의 시간적 불일치 문제를 해결하기 위해 완전한 행동 사전 정보를 활용한 복구 및 재샘플링 증강 기법을 제시한다.
Recovering Complete Actions for Cross-Dataset Skeleton Action Recognition: A Recover-and-Resample Augmentation Framework
This research paper proposes a novel "recover-and-resample" augmentation framework to address the challenge of cross-dataset skeleton action recognition, improving generalizability by leveraging a "complete action prior" to generate more comprehensive training data.
Self-Supervised Skeleton-Based Human Action Recognition under Occlusions: A Two-Stage Approach
This research paper introduces a novel approach to improve the performance of self-supervised skeleton-based human action recognition models in the presence of occlusions, a common challenge in real-world applications.
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