Low Light Object Detection using DL

Abstract—Low-light conditions significantly degrade imagequality, leading to poor visibility, noise, and loss of importantfeatures, which negatively affect object detection performance.Conventional object detection models trained on well-lit imagesfail to generalize effectively in dark environments. This paperpresents a deep learning–based low-light object detection(LLOD) framework that integrates image enhancement withYOLOv8-based object detection. The proposed system firstenhances low-light images using contrast and illuminationenhancement techniques and then performs object detection onthe enhanced images. The ExDark dataset is used forexperimentation, and annotations are created using Roboflow tosupport supervised training. Experimental results demonstratethat image enhancement significantly improves detectionaccuracy, precision, and robustness under low-illuminationconditions. The proposed approach is suitable for real-world applications such as surveillance, traffic monitoring, and night-time vision systems.