ハワイの熱帯気候には、固有種のロウルの木（Pritchardia属）をはじめとする様々なヤシの木が生息しています。これらのそびえ立つヤシの木は、夕焼け空を背景に雄大なシルエットを描き出します。
Tradewind Technology®を採用したこのシャツは、吸湿発散性に優れ速乾性があるため、湿気の多い気候に最適です。伸縮性、しわになりにくさ、UPF（紫外線保護指数）による日焼け防止効果も強化されています。
-
ナイロン92％／スパンデックス8％
-
Tradewind Technology®
-
ボタンダウンカラー
-
柄合わせされたポケット
-
ウフブルーのアクセントが施されたレーザー彫刻ボタン
-
洗濯機洗い可
Abstract\nTo improve object detection speed and accuracy for traffic sign detection tasks, this paper proposes a lightweight traffic sign detection method called YOLOv5-light based on YOLOv5. First, we replace the backbone network in YOLOv5 with a lightweight backbone network to reduce the model size and improve the inference speed. Then, for the problems of small traffic sign targets and fuzzy features, we propose a novel feature fusion method that combines high-resolution features and low-resolution features to improve the detection accuracy of small targets. Finally, we use the improved K-means++ algorithm to cluster the aspect ratio of the bounding boxes to improve the detection accuracy for different sizes of traffic sign targets. Experiments show that compared with the original YOLOv5 model, the YOLOv5-light model proposed in this paper reduces the model size by 84.7%, improves the inference speed by 35.7% (25.1 FPS on the NVIDIA Jetson Nano), and improves the mAP of 94.6% on the Chinese Traffic Sign Dataset (CTSD) and 93.3% on the Tsinghua-Tencent 100K (TT100K) dataset. The detection speed and accuracy of the YOLOv5-light model are better than other mainstream object detection models.
", "text_content": "To improve object detection speed and accuracy for traffic sign detection tasks, this paper proposes a lightweight traffic sign detection method called YOLOv5-light based on YOLOv5. First, we replace the backbone network in YOLOv5 with a lightweight backbone network to reduce the model size and improve the inference speed. Then, for the problems of small traffic sign targets and fuzzy features, we propose a novel feature fusion method that combines high-resolution features and low-resolution features to improve the detection accuracy of small targets. Finally, we use the improved K-means++ algorithm to cluster the aspect ratio of the bounding boxes to improve the detection accuracy for different sizes of traffic sign targets. Experiments show that compared with the original YOLOv5 model, the YOLOv5-light model proposed in this paper reduces the model size by 84.7%, improves the inference speed by 35.7% (25.1 FPS on the NVIDIA Jetson Nano), and improves the mAP of 94.6% on the Chinese Traffic Sign Dataset (CTSD) and 93.3% on the Tsinghua-Tencent 100K (TT100K) dataset. The detection speed and accuracy of the YOLOv5-light model are better than other mainstream object detection models.", "text_width": 300, "text_x_max": 587, "text_x_min": 287, "text_y_max": 729, "text_y_min": 305}
