数据集格式：labelme格式(不包含mask文件，仅仅包含jpg图片和对应的json文件)

图片数量(jpg文件个数)：7820

标注数量(json文件个数)：7820

标注类别数：20

标注类别名称:["Graffiti","Bearing","Wetspot","Rust","Spalling","Efflorescence","Hollowareas","Weathering","Restformwork","Cavity","ExposedRebars","EJoint","Crack","Drainage","ACrack","JTape","Rockpocket","PEquipment","WConccor","object"]

Graffiti - 涂鸦

Bearing - 支座

Wetspot - 潮湿点（或潮湿斑迹）

Rust - 锈蚀

Spalling - 剥落

Efflorescence - 泛碱

Hollowareas - 空洞区域

Weathering - 风化

Restformwork - 残留模板（或剩余模板）

Cavity - 空腔

ExposedRebars - 外露钢筋

EJoint - 伸缩缝

Crack - 裂缝

Drainage - 排水

ACrack - 活动裂缝

JTape - 接头胶带

Rockpocket - 岩窝（或岩石凹槽）

PEquipment - 防护设备

WConccor - 水侵蚀混凝土

object - 物体

每个类别标注的框数：

Graffiti count = 2197

Bearing count = 1201

Wetspot count = 1657

Rust count = 14665

Spalling count = 9849

Efflorescence count = 3956

Hollowareas count = 1538

Weathering count = 4621

Restformwork count = 1019

Cavity count = 9222

ExposedRebars count = 1993

EJoint count = 506

Crack count = 3612

Drainage count = 1649

ACrack count = 423

JTape count = 1076

Rockpocket count = 616

PEquipment count = 1892

WConccor count = 407

object count = 13

使用标注工具：labelme=5.5.0

所在仓库：firc-dataset

图片分辨率：640x640

标注规则：对类别进行画多边形框polygon

重要说明：可以将数据集用labelme打开编辑，json数据集需自己转成mask或者yolo格式或者coco格式作语义分割或者实例分割

特别声明：本数据集不对训练的模型或者权重文件精度作任何保证，数据集只提供准确且合理标注

图片预览：

标注例子：