Tag: action recognition

Researchers developed a multi-stream graph convolutional network and dataset to classify 13 martial arts categories on edge hardware, demonstrating development ...

Researchers have introduced AMST, an attention-guided spatio-temporal masking framework that enhances self-supervised skeleton action recognition across multiple benchmark datasets.

A Visual Transformer-based algorithm reported in Discover Artificial Intelligence assigns technical scores and grades to gymnastics, jumping, and racquet-sport movements ...

By feeding neural networks only optical flow and motion history images instead of raw video, researchers at TU Delft achieved ...

A deep learning pipeline combining YOLOv5 detection, skeleton extraction and neural network classification can recognize falls, collisions and shelf climbing ...

Researchers have developed TRACE, a self-supervised framework that teaches video AI models to reason about interventions and counterfactuals rather than ...

A team of researchers in Taiwan has built an artificial intelligence system that watches people perform cardiopulmonary resuscitation and tells ...