without gt_map

[Silentbarber]

Hello, I am very interested in your evaluation work, but if I don't have a gt_map, only a lio map, how should I evaluate it? Can you give me some guidance

[JokerJohn]

Hi, @Silentbarber
Thank you for your interest in our evaluation work.

Metrics like MME (Mean Map Entropy) or MPV (Mean Plane Variance) can be used to evaluate a map without requiring ground truth point clouds. Their underlying principles are quite straightforward: from an information-theoretic perspective, if the information content in a map is higher, the map's entropy will be lower, indicating that points in the map are more locally clustered, which implies better local continuity.

However, as the formula shows, these metrics only consider the local characteristics of points (using only the covariance matrix). Therefore, they cannot provide a fair evaluation when changes occur at a different scale or point cloud density. For example, comparing a map generated by LIO with one optimized through loop closure. There is a brief explanation of this in the experiments section of my paper.

[Silentbarber]

您好，感谢您对我们的评估工作感兴趣。

**MME（平均地图熵）**或 MPV（平均平面方差）等指标可用于评估地图，而无需地面实况点云。它们的基本原理非常简单：从信息论的角度来看，如果地图中的信息内容较高，则地图的熵将较低，这表明地图中的点更局部聚集，这意味着更好的局部连续性。

但是，如公式所示，这些指标仅考虑点的局部特征（仅使用协方差矩阵）。因此，当变化发生在不同的比例或点云密度时，它们无法提供公平的评估。例如，将 LIO 生成的 map 与优化的 through loop closure 进行比较。在我论文的实验部分对此进行了简要解释。

thank you！Wishing the paper a smooth progress！