Motorcycle Night Ride

摩托车骑行数据集是从YouTube上可用的开放媒体获取的200帧的集合，以实现对对象检测和/或以移动性为中心的AI解决方案的测试-特别是在计算机视觉驱动的摩托车头盔或其他发明途径上。
使用了六个标准类，分别是: 不可驾驶，道路，拉线，我的自行车，骑手，可移动
可移动的表示移动的物体，例如车辆，人等，不可驱动的表示无法骑行的区域。其他类是不言自明的，包括道路、车道 (包括反射器) 、骑手，当然还有自行车本身。
我们已经使用SuperAnnotate的像素编辑器作为语义分割的工具。它的工作原理是栅格逻辑，而不是矢量逻辑。导出包括COCO格式。我们已经预先打包了包含融合图像的数据集。

The Motorcycling Dataset is a collection of 200 frames sourced from open media available on YouTube, designed for testing object detection and/or mobility-centric AI solutions — particularly for computer vision-driven motorcycle helmet-related or other innovative research applications.

Six standard annotation classes are utilized, specifically: "Non-drivable", "Road", "Wire Lines", "My Bicycle", "Rider", and "Movable". "Movable" refers to moving objects such as vehicles, pedestrians, etc., while "Non-drivable" denotes areas that cannot be ridden on. The remaining classes are self-explanatory, including Road, Road Lanes (including reflectors), Rider, and of course the Bicycle itself.

We employed SuperAnnotate's Pixel Editor as the semantic segmentation annotation tool, which operates on raster logic rather than vector logic. Dataset exports are provided in COCO format. We have pre-packaged the dataset with fused images included.

搜集汇总

背景与挑战

背景概述

该数据集是一个包含200帧图像的摩托车夜间骑行数据集，源自YouTube开放媒体，用于测试对象检测和以移动性为中心的AI解决方案，特别是计算机视觉驱动的摩托车头盔等应用。数据采用语义分割标注，包含六个标准类别，并使用SuperAnnotate工具和COCO格式导出。