A systematic review of deep learning methods for low-light image enhancement and object detection

Object detection in low-light conditions presents significant challenges due to poor contrast, elevated noise levels, and reduced visibility, all of which hinder the effectiveness of traditional vision models. Over the years, object detection in low-light conditions has evolved from its preliminary theories, and it has embraced new improvements from Artificial Intelligence in its several concepts. This paper intends to review various taxonomies, techniques, and advancements of object detection and their subsequent enhancement strategies for low-light scenarios from inception to the current state of the art. It examines and classifies both conventional (Histogram Equalization) and advanced deep learning approaches (MSR-Net, LEGAN, R2RNet, and Retinex-based frameworks) from several key perspectives, ranging from identifying promising technologies, challenges, and opportunities that are unique to low-light object detection. This paper offers a crisp perspective on current advancements, uncovers key limitations, and outlines promising directions for future research for improving the reliability and performance of object detection systems in low-illumination environments.