YOLOv8++ with Weights Pruning for Road Object Detection in Rainy Environment
Keywords: Autonomous Driving, Object Detection, YOLOv8, Weight Pruning
Abstract: Object detection on roadways is crucial for autonomous driving and advanced driver assistance systems. However, adverse weather conditions, particularly rain, significantly degrade the performance of these systems. This paper presents a novel approach to enhance road object detection in rainy weather scenarios by applying a modified YOLOv8 model. The proposed YOLOv8++ model includes specialized data augmentation techniques to simulate rainy conditions, adjustments in the network architecture to improve robustness against rain-induced noise, and optimized training strategies to enhance model performance. The study leverages BDD100K, Cityscapes and DAWN-Rainy datasets consisting of various road scenarios under different intensities of rain. We systematically augment these datasets to ensure the model learns to identify objects obscured by rain streaks and reflections. Our YOLOv8++ model introduces enhancements in the feature extraction layers, enabling better handling of occlusions and reduced visibility. Extensive experiments demonstrate that our model outperforms the baseline YOLOv8 and other state-of-the-art object detection models in terms of mean Average Precision (mAP) under rainy conditions. Additionally, to ensure the model's efficiency and suitability for real-time applications, we apply a network pruning technique, which reduces the model size and computational requirements without sacrificing performance. This research contributes to the field of autonomous driving by providing a more reliable object detection system for adverse weather conditions, enhancing overall road safety.
Submission Number: 35
Loading