Active-Passive SimStereo

Active-Passive SimStereo数据集由西澳大利亚大学创建，包含515对图像，用于评估立体匹配算法的性能。数据集提供主动和被动立体图像，以及相应的完美地面实况视差，允许比较在被动或主动立体匹配中使用的立体视觉方法的相对性能。该数据集通过使用基于物理的渲染引擎生成的计算机生成图像来创建，旨在解决深度学习模型在主动立体视觉中的泛化能力问题，特别是在自动驾驶等领域的应用。

The Active-Passive SimStereo Dataset was developed by the University of Western Australia. It consists of 515 image pairs for evaluating the performance of stereo matching algorithms. The dataset provides active and passive stereo images along with their corresponding perfect ground-truth disparity maps, enabling comparative analysis of the relative performance of stereo vision methods used in either passive or active stereo matching. Constructed using computer-generated images produced by a physics-based rendering engine, this dataset aims to address the generalization capability issues of deep learning models in active stereo vision, particularly for applications in fields such as autonomous driving.