Improving Object Detection Performance Through Selective Low-Light Enhancement

This paper proposes a selective low-light enhancement algorithm and integrated NMS (Non-Maximum Suppression) operation to improve the accuracy and performance of object detection in low-light environments. The method involves selectively enhancing low-light images by applying the CLHAE (Contrast Limited Adaptive Histogram Equalization) algorithm to generate improved images. The improved and original images are then simultaneously fed into the object detection network, and the NMS is applied to remove redundant detections and obtain the final results. The proposed approach is evaluated on the ExDark dataset, demonstrating superior performance compared to existing methods.