Tie-Hua Du

Automated Vision-Based Sensing of Paediatric Pain Using 2D-Facial Landmark Trajectories Derived from 3D Geometric Normalisation and a Spatial-Temporal Attention Long Short-Term Memory (STA-LSTM) Network

Highlights What are the main findings? A vision-based sensing pipeline using 2D trajectories derived from 3D-normalised landmarks and a STA-LSTM network can accurately distinguish observer-labelled pain from no-pain states in children (AUC 0.994, accuracy 97.15%) The system operates on landmark coordinates rather than...