GRA: Graph Representation Alignment for Semi-Supervised Action Recognition.

Graph convolutional networks (GCNs) have emerged as a powerful tool for action recognition, leveraging skeletal graphs to encapsulate human motion. Despite their efficacy, a significant challenge remains the dependency on huge labeled datasets. Acquiring such datasets is often prohibitive, and the frequent occurrence of incomplete skeleton data, typified by absent joints and frames, complicates the testing phase.

Facile Photoresponsive Actuators Based on Ferrocene-Doped Poly(butyl methacrylate).

This paper presents facile photoresponsive actuators comprising ferrocene as a guest chromophore and poly(butyl methacrylate) (PBMA) as a host matrix. The ferrocene-doped PBMA film exhibits mechanical expansion and contraction when a 445 nm laser is turned on and off, respectively. The photoresponsive film is attached by a commercially available acetylcellulose adhesive tape, which exhibits a bending motion that is controlled by turning the laser on and off.