jibingyang111/PersonDataset

PersonDataset是一个专门为密集人群场景深度估计任务设计的合成数据集，使用Unity引擎创建。该数据集包含左右图像对以及与左图像对应的地面真实深度图。相机先在左侧位置捕获左图像，然后水平移动一米到右侧捕获右图像。相机的焦距为600像素。深度图可以通过公式转换为视差图。数据集的训练集与测试集的比例为8:2，共有8000个图像对，每个图像对包括一个左图像、一个右图像和对应的深度图。每个图像的分辨率为906×415。

PersonDataset is a synthetic dataset designed for depth estimation tasks in dense crowd scenarios, created using the Unity engine. The dataset includes pairs of left and right images, along with the ground truth depth maps corresponding to the left images. The camera captures the left image from the left position, then moves one meter horizontally to the right to capture the right image. The cameras focal length is 600 pixels. Depth maps can be converted to disparity maps using the formula disp = (focal length × 1) / depth. The training to testing set ratio is 8:2, with a total of 8,000 image pairs. Each pair contains a left image, a right image, and the corresponding depth map. The resolution of each image is 906×415.