Detecting accidents in night-time videos presents unique challenges due to low visibility, varying lighting conditions, and motion blur, which can obscure crucial visual cues. To address these issues, we propose a detection transformer-based model specifically trained for detection of accidents during night in traffic videos. Our model leverages the power of attention-based architectures to effectively identify and localize objects under poor lighting and low visibility. Extensive experiments on a diverse set of night videos demonstrate the superior performance of our approach, achieving an Average Precision (AP) of 0.982 and an Average Recall (AR) of 0.910. These results highlight the efficacy of our experiments in accurately identifying accidents, even under the most challenging visibility conditions.