语义分割-车道线

数据集包括16000张可以直接用于训练的车载影像数据，采用分割连通域标注方法，对这些图片数据标注了虚车道线、实车道线和斑马线的区域和类别，其中标注数据以灰度图的方式存储。 标注数据是与原图尺寸相同的单通道灰度图，其中背景像素的灰度值为0，不同类别的目标像素分别为不同的灰度值。实车道线、虚车道线和斑马线类别编号分别为0、1和2This dataset includes 16,000 pieces of directly trainable in-vehicle image data. The segmented connected component annotation method is adopted to label the regions and categories of dashed lane lines, solid lane lines and zebra crossings for these images. The annotated data are stored as single-channel grayscale images with the same size as the original images, where the grayscale value of background pixels is 0, and target pixels of different categories have different grayscale values respectively. The category numbers of solid lane lines, dashed lane lines and zebra crossings are 0, 1 and 2 respectively.

搜集汇总

背景与挑战

背景概述

该数据集包含16000张车载影像，专门用于车道线语义分割任务，标注了虚车道线、实车道线和斑马线三类目标，采用灰度图存储标注信息，适用于自动驾驶或道路场景分析模型的训练。