Spatio-temporal hard attention learning for skeleton-based activity recognition

Article Computer Science, Artificial Intelligence

Spatio-temporal attention on manifold space for 3D human action recognition

Chongyang Ding et al.

Summary: In this study, a novel approach for skeleton-based action recognition is proposed, incorporating spatial and temporal attention mechanisms on Lie groups for 3D human action recognition. By introducing a generalized attention mechanism on Lie groups, the scope of attention is extended from traditional Euclidean space to manifold space, allowing for the identification of significant spatial features and temporal stages.

APPLIED INTELLIGENCE (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

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

Wei Peng et al.

Summary: Graph Convolutional Network (GCN) is successfully applied to skeleton-based action recognition, however, lacking pooling operations leads to flat architectures, Tripool offers a solution for this issue.

PATTERN RECOGNITION (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

Host-Parasite: Graph LSTM-in-LSTM for Group Activity Recognition

Xiangbo Shu et al.

Summary: This article introduces a novel Graph LSTM-in-LSTM (GLIL) method for group activity recognition by modeling the person-level actions and group-level activity simultaneously, aiming to tackle the problem of group activity recognition in the multiple-person scene.

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields

Zhe Cao et al.

Summary: This work presents a real-time approach to detect the 2D pose of multiple people in an image, achieving high accuracy and real-time performance regardless of the number of people in the image. By refining PAFs separately, both runtime performance and accuracy are significantly improved. Additionally, the first combined body and foot keypoint detector reduces inference time and maintains accuracy compared to running them sequentially.

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE (2021)

Add to Collection

Article Computer Science, Artificial Intelligence