# YOLO-Night: Lighting the Path for Autonomous Vehicles with Robust Nighttime Perception

## TL;DR

YOLO-Night is a new object detection framework that improves nighttime perception for autonomous vehicles by adapting the YOLO architecture to low-light conditions.

## Contribution

YOLO-Night introduces architectural adaptations like feature conditioning and multi-scale fusion for robust nighttime object detection.

## Key findings

- YOLO-Night outperformed lightweight YOLO baselines in precision and mAP on the NightCity dataset.
- YOLO-Night achieved +14.3% precision, +12.4% recall, and +10.4% mAP@50 improvements over YOLO11n under nighttime conditions.
- YOLO-Night maintains real-time inference with moderate computational overhead, making it suitable for real-world deployment.

## Abstract

What are the main findings?
A nighttime-oriented YOLO framework (YOLO-Night) is proposed, integrating feature conditioning, adaptive receptive fields, and staged multi-scale fusion to improve detection robustness under low-illumination conditions.YOLO-Night achieved substantially higher precision and mAP on the NightCity dataset than lightweight YOLO baselines and nighttime-oriented detectors while maintaining real-time inference with moderate computational overhead.

A nighttime-oriented YOLO framework (YOLO-Night) is proposed, integrating feature conditioning, adaptive receptive fields, and staged multi-scale fusion to improve detection robustness under low-illumination conditions.

YOLO-Night achieved substantially higher precision and mAP on the NightCity dataset than lightweight YOLO baselines and nighttime-oriented detectors while maintaining real-time inference with moderate computational overhead.

What are the implications of the main findings?
The results demonstrate that architectural adaptation within the detection pipeline is more effective than standalone image enhancement for nighttime autonomous driving perception.YOLO-Night provides a practical accuracy–efficiency trade-off suitable for real-world deployment in autonomous vehicles and Advanced Driver Assistance Systems (ADAS).

The results demonstrate that architectural adaptation within the detection pipeline is more effective than standalone image enhancement for nighttime autonomous driving perception.

YOLO-Night provides a practical accuracy–efficiency trade-off suitable for real-world deployment in autonomous vehicles and Advanced Driver Assistance Systems (ADAS).

Despite substantial progress in visual perception, object detection systems for autonomous driving still exhibit pronounced performance degradation in nighttime and low-light conditions, where reduced signal-to-noise ratio, blurred object boundaries, and scale ambiguity challenge reliable recognition. Existing YOLO-based detectors, primarily optimized for daytime imagery, struggle to maintain robustness under such adverse illumination. To address these issues, we propose YOLO-Night, a nighttime-oriented object detection framework that enhances the YOLO11 architecture through a structured integration of contrast enhancement, adaptive receptive field modeling, and multi-scale feature fusion. The framework incorporates a feature-level enhancement mechanism to improve low-contrast representations, employs depthwise switchable atrous convolution to dynamically adapt receptive fields for blurred and small objects, and introduces a multi-scale convolutional block to strengthen feature extraction under severe illumination degradation. In addition, a staged feature fusion strategy with an auxiliary low-level detection head was adopted to mitigate semantic misalignment across feature scales. Extensive experiments on the NightCity dataset demonstrated that YOLO-Night consistently outperformed the YOLO11n baseline, achieving improvements of +14.3% precision, +12.4% recall, and +10.4% mAP@50 under nighttime conditions while maintaining real-time inference capability. These results indicate that targeted architectural adaptations can substantially improve object detection robustness in low-light driving scenarios.