Researchers have developed FUSE (Frame-Unified Stress Estimation), a framework for detecting stress from facial video without breaking it down into short clips. Instead of dividing the recording into windows, FUSE processes all frames as one unified input. This approach achieves high accuracy and can handle full-length recordings within a single architecture. Experiments on a 58-subject dataset show that FUSE outperforms other methods in certain configurations, with test accuracy reaching up to 69.44%. The framework's efficiency varies depending on the temporal stride used. --- Why it matters: This matters because it shows that existing approaches to stress detection from facial video may not be necessary, and a more unified approach can achieve similar or better results. This could simplify the development of affect monitoring systems. Source: https://arxiv.org/abs/2608.10442