bshepp/residuals-fingerprints

RESIDUALS数据集包含LiDAR DEM（数字高程模型）残差指纹，由39,716个残差图像组成。这些图像是通过对俄亥俄州费尔菲尔德县的单一LiDAR-derived DEM（1500×375，分辨率为3.33英尺/像素）应用593种不同的分解配置和25种上采样方法生成的。每行数据包含一个256×256的PNG残差图像（使用标准`RdBu_r`色彩映射和99百分位对称裁剪渲染）、生成该图像的算法和参数、一个40维特征向量以及预计算的2D/3D UMAP坐标。该数据集适用于算法分类基准测试、算法指纹识别/信号处理取证等任务。所有残差图像均来自同一地形，为研究不同算法的保留与破坏特性提供了干净的实验基础。数据集还包含详细的模式描述、源数据信息、分割信息和使用示例。

The RESIDUALS dataset consists of LiDAR DEM residual fingerprints, comprising 39,716 residual images. These images were generated by applying 593 distinct decomposition configurations and 25 upsampling methods to a single LiDAR-derived Digital Elevation Model (DEM) in Fairfield County, Ohio (1500×375 at 3.33 ft/px resolution). Each row pairs a 256×256 PNG of the residual (rendered with the standard `RdBu_r` colormap and 99th-percentile symmetric clipping) with the algorithm and parameters that produced it, a 40-dimensional signature vector, and pre-computed 2D/3D UMAP coordinates. The dataset is designed for tasks such as algorithm classification benchmarking, algorithm fingerprinting/signal-processing forensics, and studying the effects of different algorithms on the same underlying terrain. It includes detailed schema descriptions, source data information, splits, and usage examples.