Evaluating the Performance of Object Detection Algorithm in Low Light

In this crucial area of computer vision, this study tackles the difficulties of identifying objects in dim illumination. As deep learning is advancing quickly and sophisticated CNN designs appear, this research investigates and compares four modern and advanced algorithms—Faster RCNN, Mask RCNN, SPP Net, and Mobile Net—tailored for efficient low-light object detection. Utilizing a comprehensive NOD dataset, captured under varying low-light scenarios, this research work evaluates the algorithms based on precision, recall, F1 score, and accuracy. The findings reveal the Mobile Net’s exceptional performance in balancing accuracy and efficiency, which will make the system robust for real-time applications. This research not only benchmarks the algorithms quantitatively but also provides qualitative insights into their computational requirements, yielding valuable considerations for selecting suitable models in challenging illumination scenarios.