Researchers have developed a lightweight U-Net model using depthwise separable convolutions. This model enables real-time, high-accuracy human activity recognition on wearable devices with minimal ...