Driver Monitoring System has become an integral part of modern vehicle safety systems. However, traditional monitoring systems face limitations in low-light conditions, especially at night or in weak light, making it challenging to accurately capture the driver's state. This paper proposes a U-shaped structure for no-reference low-light image enhancement, introducing channel and spatial attention mechanisms into the U-shaped structure. This lightweight neural network adjusts the brightness of individual pixels in images. Qualitative and quantitative experiments demonstrate that U-ATTENTION-NET achieves excellent low-light image enhancement results, with a significantly reduced number of parameters compared to traditional low-light image methods. U-ATTENTION-NET also boasts high inference speed, capable of meeting real-time frame-by-frame detection at 30 frames per second for camera applications.