Deep Learning Based Night Time Vehicle Detection for Autonomous Cars using Generative Adversarial Network

Vehicular detection is an important research area in the field of target detection. The implementation of accurate vehicle recognition models for both day and night conditions will make future automatic vehicle technologies and driver assistance systems more reliable. Our research proposes a novel approach for night vehicle detection. The system is made up of two modules: one for image translation and one for vehicle detection. When a nighttime image is given, the picture translation module uses the Generative Adversarial as a basic network to produce daytime images as an output. Faster R-CNN and YOLOv5 algorithms are used during the vehicle detection phase to extract daytime features and identify vehicles. The YOLOv5 technology achieves a detection accuracy of 96.75%.