BDD-A-C

该数据集名为BDD-A-C，是DriverAttention-C数据集的一个子集，专注于在多样化的城市和乡村道路驾驶场景中进行注意力预测。该子集包含了与实际驾驶条件相关的常见损坏情况。其规模包括6,817张测试图像，旨在为自动驾驶系统执行注意力预测任务。This dataset, named BDD-A-C, is a subset of the DriverAttention-C dataset, focusing on attention prediction across diverse urban and rural road driving scenarios. This subset includes common corruptions associated with real-world driving conditions. It consists of 6,817 test images and is designed to support attention prediction tasks for autonomous driving systems.

搜集汇总

背景与挑战

背景概述

BDD-A-C是DriverAttention-C基准的合成子集，包含超过11.5万张图像，涵盖噪声、模糊、天气等多种腐败类别，用于评估自动驾驶中注意力预测模型的鲁棒性。该数据集通过手动重新标注确保地面真实性，旨在解决真实驾驶场景中的腐败和中心偏差挑战。