Comparative Study of YOLOv11, SSD, and Fast RNN for Nighttime Face Detection on the DarkFace Dataset

Object detection in low-light scenarios has always been a challenge. The detector, which was been trained in a well-illuminated dataset, presents lower performance in low-light scenarios. This is mainly due to factors such as low image visibility and large noise interference that affect the clarity and distinguishability of the target. To study the performance of different detection models in the task of night face detection, this paper selects three typical object detection models, YOLOv11, Fast Region-based Convolutional Network (Faster R-CNN), and Single Shot MultiBox Detector (SSD), based on the DarkFace night face dataset for performance evaluation and comparison. In the experiment, the DarkFace dataset is used for testing, and [email protected] is used as the main evaluation indicator. The results show that You Only Look Once version 101 (YOLOv11) is significantly better than Faster R-CNN and SSD models in terms of accuracy and F1-recall and has stronger night object detection capabilities, which means that YOLOv11 has better robustness and detection performance in low-light environments and can be used as an effective solution for night face recognition tasks.