Benchmark data for: Machine Learning for geospatial vector data classification

Benchmark data for paper "Deep Learning for Classification Tasks on Geospatial Vector Polygons". Core of the data is in the six numpy zip files. Each numpy zip contains the original WKT geometries as zlib compressed blobs, variable and fixed length geometry vectors, fourier descriptors, and a class dictionary.

The zlib compressed wkt strings can be decompressed with
<code>
<br>
import numpy as np<br>
import zlib<br>
<br>
loaded = np.load('archaeology_train_v8.npz')<br>
wkts_zipped = loaded['wkts_zlib_compressed']<br>
for wkt_zipped in wkts_zipped:<br>
&nbsp wkt = str.decode(zlib.decompress(wkt_zipped))<br>
</code>

本数据集为论文《"Deep Learning for Classification Tasks on Geospatial Vector Polygons"》（面向地理空间矢量多边形分类任务的深度学习）的基准测试数据集。数据集的核心内容存放于6个NumPy压缩存档文件（.npz格式）中。每个此类压缩文件均包含以zlib压缩二进制块形式存储的原始WKT（Well-Known Text）几何数据、可变长度与固定长度几何向量、傅里叶描述子（Fourier Descriptors）以及类别字典。

可通过以下代码解压该zlib压缩的WKT字符串：
import numpy as np
import zlib

loaded = np.load('archaeology_train_v8.npz')
wkts_zipped = loaded['wkts_zlib_compressed']
for wkt_zipped in wkts_zipped:
wkt = str.decode(zlib.decompress(wkt_zipped))