Enhancing Object Detection in Low Light Environments Using Image Enhancement Techniques and YOLO Architectures

This study analyzes how YOLO structures and image enhancement techniques can improve object detection in low light conditions. In order to overcome the difficulties caused by glare, noise, and dim lighting, the study makes use of the Exclusively Dark (ExDark) dataset, which covers a wide variety of low light situations. The research showed significant enhancements in object detection precision through the utilization of Convolutional Neural Networks (CNN) and pre-processing techniques. A comparative analysis of YOLO versions v7, v8, and v9 is included in the paper, highlighting the superior performance of YOLO v7 when combined with Enlighten GAN, achieving a mAP 0.5 of 73.7%. The results demonstrate how useful preprocessing methods are when used with YOLO algorithms to provide reliable object detection in difficult low-light situations.