SeasonDepth
收藏资源简介:
SeasonDepth是一个专为跨季节单目深度预测设计的新数据集,由卡内基梅隆大学创建。该数据集包含17225张图像,涵盖多种环境条件,如不同的光照和季节变化,旨在评估深度估计算法在不同环境下的性能。数据集通过结构从运动(SfM)和多视图立体(MVS)管道构建,支持对代表性的开源监督和自监督深度预测方法进行基准测试。SeasonDepth的应用领域包括自动驾驶和移动机器人的长期视觉感知,旨在解决学习型算法在多变环境中的泛化问题。
SeasonDepth is a novel dataset designed for cross-season monocular depth prediction, created by Carnegie Mellon University. This dataset contains 17,225 images covering diverse environmental conditions including varying lighting and seasonal changes, which aims to evaluate the performance of depth estimation algorithms across different environments. Constructed using Structure from Motion (SfM) and Multi-View Stereo (MVS) pipelines, it supports benchmarking representative open-source supervised and self-supervised depth prediction methods. The application fields of SeasonDepth cover long-term visual perception for autonomous driving and mobile robotics, aiming to address the generalization challenge of learning-based algorithms in variable environments.




