CVPR 2018 WAD Video Segmentation Challenge

在本竞赛中，预测在车载摄像机视图中出现的不同可移动对象的分割。该数据集包含大量分割后的原始驾驶图像。有多个标签。在本次比赛中，评估了七个不同的实例级别注释，分别是汽车，摩托车，自行车，行人，卡车，公共汽车和三轮车。当标记无法区分边界时，将对相应的组（例如汽车组和自行车组）进行注释。目前尚未评估这些组。

In this competition, the task is to perform segmentation of various movable objects appearing in in-vehicle camera views. This dataset contains a large number of raw driving images with segmentation annotations and includes multiple label categories. Seven distinct instance-level annotations are evaluated in this competition, namely cars, motorcycles, bicycles, pedestrians, trucks, buses, and tricycles. When boundaries between certain categories cannot be distinguished, corresponding groups (e.g., the car group and bicycle group) will be annotated. Currently, these group-level annotations have not been evaluated.

搜集汇总

背景与挑战

背景概述

该数据集源自CVPR 2018 WAD视频分割挑战赛，专注于从车载摄像机图像中预测不同可移动对象的分割。它包含大量已标注的驾驶场景图像，涵盖汽车、摩托车、自行车等多个类别，用于实例级分割任务。