The paper "TSD-YOLO: Small Traffic Sign Detection based on Improved YOLO v8" addresses the challenges of detecting small traffic signs in complex environments, particularly in autonomous driving scenarios. The authors propose an enhanced version of the YOLO v8 model, focusing on improving detection accuracy and speed, especially for small and occluded objects. The key contributions include: 1. **Select Kernel (SK) Attention Mechanism**: This mechanism dynamically adjusts the weights of feature channels to enhance the focus on critical features, improving detection in complex backgrounds and occlusions. 2. **Space-to-Depth (SPD) Module**: Integrated into the backbone network, this module distributes spatial information into depth channels, accelerating feature extraction and expanding the receptive field, thereby enhancing the model's ability to detect traffic signs across various scales. 3. **WLOv3 Loss Function**: Used as the bounding box loss function, this function assigns different weights to samples based on their occurrence frequencies, improving the model's ability to handle objects of various sizes and frequencies, particularly small or rare traffic signs. The paper also reviews related work, including the evolution of YOLO versions and other traffic sign detection methods, and presents experimental results on the CCTSDB and TT100K datasets. Ablation experiments demonstrate the effectiveness of the proposed modules, showing significant improvements in mAP50 metrics. The proposed method outperforms other mainstream single-stage detection models, achieving higher mAP50 values and better recall and accuracy in detecting small objects. The authors conclude by highlighting the model's enhanced performance in complex traffic scenes and future directions for improving robustness and inference speed.The paper "TSD-YOLO: Small Traffic Sign Detection based on Improved YOLO v8" addresses the challenges of detecting small traffic signs in complex environments, particularly in autonomous driving scenarios. The authors propose an enhanced version of the YOLO v8 model, focusing on improving detection accuracy and speed, especially for small and occluded objects. The key contributions include: 1. **Select Kernel (SK) Attention Mechanism**: This mechanism dynamically adjusts the weights of feature channels to enhance the focus on critical features, improving detection in complex backgrounds and occlusions. 2. **Space-to-Depth (SPD) Module**: Integrated into the backbone network, this module distributes spatial information into depth channels, accelerating feature extraction and expanding the receptive field, thereby enhancing the model's ability to detect traffic signs across various scales. 3. **WLOv3 Loss Function**: Used as the bounding box loss function, this function assigns different weights to samples based on their occurrence frequencies, improving the model's ability to handle objects of various sizes and frequencies, particularly small or rare traffic signs. The paper also reviews related work, including the evolution of YOLO versions and other traffic sign detection methods, and presents experimental results on the CCTSDB and TT100K datasets. Ablation experiments demonstrate the effectiveness of the proposed modules, showing significant improvements in mAP50 metrics. The proposed method outperforms other mainstream single-stage detection models, achieving higher mAP50 values and better recall and accuracy in detecting small objects. The authors conclude by highlighting the model's enhanced performance in complex traffic scenes and future directions for improving robustness and inference speed.