Charlotte Area Traffic Light Dataset

We present a new traffic light object detection dataset of over 4,000 images labeled from dashcam video in Charlotte, NC. The dataset captures challenging weather and lighting, with daytime, nighttime, and rainy conditions. 7 annotation classes allow benchmarking of detection models like YOLOv5. Performance degrades at night and in rain versus daytime conditions. This geographically targeted dataset exposes region-specific factors impacting traffic light perception. It serves as a benchmark tailored to Charlotte to improve robustness across settings.