ECP New York 2011

by No License

ECP New York 2011

The ECP New York dataset contains 10 manually segmented buildings from New York City, USA. Segmentation evaluating using Dice coefficient is calculated for the corresponding 1D horizontal and vertical segmentations as well as the complete 2D binary segmentation.The topology/structure of the segmentation is evaluated by comparison of the correctness of the retrieved number of patterns. The measure is defined such that [Teboul2011]: A pattern is supposed to be detected if its label in the Markov chain has a non-zero size: w6= 0. Therefore, we can count the number of detected patterns nd and compare it to the ground truth ngt and compute the topological precision.

Dataset Attributes

Label SVG
CategoriesProcedural, Reconstruction, Urban, Newyork