Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
4
4
image
stringlengths
15
15
width
int32
630
27.9k
height
int32
723
31.1k
boxes
listlengths
2
6.89k
polygons
listlengths
2
6.89k
labels
listlengths
2
6.89k
attributes
listlengths
2
6.89k
relations
dict
0000
images/0000.png
3,584
2,048
[ [ 3109.5, 394.5, 3142.5, 583.5 ], [ 3109.5, 609, 3142.5, 867 ], [ 2182.5, 882, 2377.5, 924 ], [ 2014.5, 870, 2173.5, 912 ], [ 1851, 856.5, 1881, 997.5 ], [ 2016, 927, 2094, 945 ], [ 2172, ...
[ [ [ 3143, 585 ], [ 3113, 586 ], [ 3110, 396 ], [ 3143, 396 ] ], [ [ 3143, 869 ], [ 3114, 869 ], [ 3112, 610 ], [ 3145, 611 ] ], [ [ 2183, ...
[ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2...
[ [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ ...
{ "subject_index": [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 293, 25, 26, 59, 60, 71, 290, 291, 292, 73, 74, 75, 76,...
0001
images/0001.png
6,400
6,912
[ [ 4602, 4206, 4680, 4278 ], [ 4689, 4170, 4773, 4248 ], [ 4695, 4420.5, 4743, 4465.5 ], [ 4752, 4309.5, 4800, 4354.5 ], [ 4263, 4497, 4341, 4569 ], [ 4362, 4473, 4410, 4515 ], [ 4387.5, 45...
[ [ [ 4682, 4259 ], [ 4663, 4207 ], [ 4603, 4228 ], [ 4621, 4279 ] ], [ [ 4690, 4197 ], [ 4713, 4249 ], [ 4773, 4229 ], [ 4754, 4170 ] ], [ [ ...
[ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9...
[ [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ], [ ...
{ "subject_index": [ 7, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 0, 1, 2, 3, 4, 5, 6, 29, 30, 31, 32, 33, 36, 37, 38, 39, 40, 41, ...
0011
images/0011.png
7,936
7,424
[[1551.0,2418.0,4215.0,6138.0],[3487.5,1365.0,5992.5,4863.0],[3847.5,1108.5,6346.5,4603.5],[1345.5,2(...TRUNCATED)
[[[1552.0,2470.0],[1630.0,2418.0],[4216.0,6083.0],[4141.0,6139.0]],[[3488.0,1413.0],[3560.0,1365.0],(...TRUNCATED)
[11,11,11,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,(...TRUNCATED)
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[86,87,88,89,90,93,94,96,98,99,100,101,102,103,104,105,106,107,107,155,156,158,159,(...TRUNCATED)
0016
images/0016.png
2,560
3,072
[[1251.0,882.0,1281.0,912.0],[961.5,1021.5,994.5,1054.5],[969.0,1276.5,999.0,1309.5],[954.0,1305.0,9(...TRUNCATED)
[[[1281.0,893.0],[1269.0,913.0],[1250.0,903.0],[1263.0,882.0]],[[996.0,1035.0],[980.0,1058.0],[963.0(...TRUNCATED)
[ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 11, 13, 13, 12, 12, 12, 12, 12, 12, 14, 14, 14, 14, 14, 14, 14, 14 ]
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,17,19,20,21,19,20,21,32,38,32,26,28(...TRUNCATED)
0019
images/0019.png
5,406
7,681
[[2140.5,447.0,2203.5,501.0],[2058.0,685.5,2142.0,766.5],[2053.5,787.5,2110.5,838.5],[2016.0,852.0,2(...TRUNCATED)
[[[2202.0,459.0],[2191.0,502.0],[2140.0,489.0],[2151.0,446.0]],[[2143.0,704.0],[2125.0,770.0],[2060.(...TRUNCATED)
[9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9(...TRUNCATED)
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,23,24,25,26,28,29,30,31,36(...TRUNCATED)
0022
images/0022.png
6,041
3,971
[[4488.0,664.5,4536.0,709.5],[4488.0,724.5,4572.0,805.5],[4510.5,970.5,4609.5,1063.5],[4278.0,1125.0(...TRUNCATED)
[[[4527.0,666.0],[4538.0,699.0],[4500.0,712.0],[4489.0,678.0]],[[4551.0,726.0],[4575.0,783.0],[4511.(...TRUNCATED)
[9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9(...TRUNCATED)
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[0,1,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,3(...TRUNCATED)
0023
images/0023.png
3,293
4,083
[[627.0,1554.0,1047.0,2040.0],[2478.0,2310.0,2844.0,2754.0],[1675.5,1228.5,1864.5,1405.5],[1789.5,12(...TRUNCATED)
[[[627.0,1615.0],[979.0,1554.0],[1048.0,1976.0],[692.0,2041.0]],[[2479.0,2354.0],[2806.0,2310.0],[28(...TRUNCATED)
[15,15,16,16,16,16,16,16,16,16,16,16,16,16,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,18,18,18,19,(...TRUNCATED)
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[2,3,14,15,16,17,18,19,20,21,22,23,24,25,11,9,10,11,9,10,13,8,6,7],"object_index":[(...TRUNCATED)
0024
images/0024.png
9,781
5,469
[[1734.0,1573.5,2268.0,2086.5],[8235.0,3609.0,8745.0,4131.0],[8593.5,4362.0,8842.5,4530.0],[7954.5,4(...TRUNCATED)
[[[1735.0,1727.0],[2130.0,1573.0],[2270.0,1927.0],[1873.0,2086.0]],[[8236.0,3756.0],[8596.0,3610.0],(...TRUNCATED)
[15,15,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,17,17,17,17,17,17,17,17,17,17,17,(...TRUNCATED)
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[22,23,25,24,26,28,29,30,31,32,33,4,5,10,11,12,13,14,17,18,19,20,21,2,6,3,7,48,27,4(...TRUNCATED)
0025
images/0025.png
5,319
5,363
[[3666.0,4374.0,4218.0,4830.0],[2080.5,708.0,2695.5,1242.0],[3021.0,1233.0,3057.0,1317.0],[3058.5,12(...TRUNCATED)
[[[3666.0,4513.0],[4133.0,4376.0],[4220.0,4693.0],[3750.0,4833.0]],[[2081.0,913.0],[2581.0,708.0],[2(...TRUNCATED)
[ 15, 15, 17, 17, 17, 17, 17, 17, 17, 16, 16, 16, 16, 16, 18, 18, 18, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19 ]
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[2,3,4,5,6,7,8,9,10,11,10,12,13,2,2,3,2,2,3,5,5,5,6,6,7,5,5,5,6,6,7,9,9,10],"object(...TRUNCATED)
0030
images/0030.png
5,376
5,120
[[1690.5,477.0,1969.5,579.0],[2973.0,292.5,3087.0,337.5],[3240.0,217.5,3426.0,286.5],[3253.5,207.0,3(...TRUNCATED)
[[[1700.0,582.0],[1690.0,538.0],[1957.0,479.0],[1969.0,531.0]],[[2977.0,340.0],[2974.0,314.0],[3083.(...TRUNCATED)
[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2(...TRUNCATED)
[[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,(...TRUNCATED)
{"subject_index":[0,1,2,4,466,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,7(...TRUNCATED)
End of preview. Expand in Data Studio

STAR Scene Graph Dataset

STAR(Scene graph generaTion in lArge-size satellite imageRy)是面向大幅高分辨率卫星影像场景图生成的数据集。本仓库依据官方 SGG-ToolKit 发布的 STAR-SGG-with-attri.h5、taxonomy 和图片元数据构建。

图片作为仓库普通文件保存在 images/,Parquet 中的 image 字段是仓库相对路径,不包含图片字节,也不会自动解码为 Pillow 对象。完整仓库约 127 GB,下载前请确认磁盘空间。

Split

本仓库使用项目内固定的 H5 行号索引划分数据,而不是使用 H5 的 split 数值:

Hugging Face split 图片 对象 关系
train 771 131,470 203,889
validation 245 43,800 100,932
test 264 43,828 99,771
split 行总计 1,280 219,098 404,592

三个 split 涉及 1,269 张唯一图片。validation 与 test 共享图片 ID 662, 676, 717, 748, 823, 824, 825, 837, 873, 989, 1017;图片 ID 533, 884, 889, 1198 不属于任何公开 split。各 split 内均按 H5 行号升序排列。

字段

每行对应一张图片:

  • id:四位图片 ID。
  • image:仓库相对路径,例如 images/0000.png;路径本身不包含 split。
  • widthheight:图片的实际像素尺寸。
  • boxes:与对象平行的 [x1, y1, x2, y2] 水平框。
  • polygons:与对象平行的四点 polygon,每点为 [x, y]
  • labels:对象类别。
  • attributes:每个对象对应的 10 个官方属性槽;0 是 padding/background。
  • relations.subject_indexrelations.object_index:关系端点在当前对象数组中的零基索引。
  • relations.predicate:关系谓词类别。

boxespolygonslabelsattributes 长度相同。三个 relations 数组也彼此平行。本仓库保留 H5 的所有关系;不会执行训练代码中可选的重复关系随机采样。

原始目录中的 1273 个文件均使用 .png 扩展名;其中 1179 个实际采用 PNG 编码、94 个实际采用 JPEG 编码。发布过程通过硬链接保留官方源文件字节,不执行转码。

SGG-ToolKit 坐标约定

公开字段遵循官方 SGG-ToolKit 的 H5 读取规则:

  • boxes_2000 由中心点—宽高转换为角点框后乘以 6000 / 2000
  • segmentation_2000 同样乘以 6000 / 2000,并保留原始四点顺序。
  • H5 中的全局关系端点已转换成每张图片内部的零基对象索引。

对象、谓词和属性 taxonomy 均在索引 0 保留 __background__。H5 的前景标签已经从 1 开始,因此其前景整数编号保持不变。

下载与读取

from pathlib import Path

from datasets import load_dataset
from huggingface_hub import snapshot_download


revision = "main"
repository = Path(
    snapshot_download(
        repo_id="wliafe/STAR",
        repo_type="dataset",
        revision=revision,
    )
)
dataset = load_dataset("wliafe/STAR", revision=revision)

sample = dataset["train"][0]
image_path = repository / sample["image"]
print(sample["id"], image_path, sample["width"], sample["height"])

使用 Pillow 打开图片:

from PIL import Image


with Image.open(image_path) as image:
    image.load()
    print(image.size)

训练或评测时,snapshot_download()load_dataset() 应使用同一个完整 commit revision,以保证图片和 Parquet 标注来自同一版本。

使用限制

  • STAR 类别与谓词存在长尾分布。
  • 大尺寸遥感影像需要较多下载空间、内存和 I/O 带宽。
  • 标注可能包含遗漏、歧义或类别噪声。
  • 使用者应遵守 STAR 原始发布方的许可、引用和使用要求。

引用

@article{li2025star,
  title={STAR: A First-Ever Dataset and a Large-Scale Benchmark for Scene Graph Generation in Large-Size Satellite Imagery},
  author={Li, Yansheng and Wang, Linlin and Wang, Tingzhu and others},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  volume={47},
  number={3},
  pages={1832--1849},
  year={2025},
  doi={10.1109/TPAMI.2024.3508072}
}
Downloads last month
709