当前位置：首页 > news >正文

道路病害基础设施损坏标线标志牌褪色检测数据集4083张VOC+YOLO

news 2026/8/4 3:27:24

数据集格式：Pascal VOC格式+YOLO格式(不包含分割路径的txt文件，仅仅包含jpg图片以及对应的VOC格式xml文件和yolo格式txt文件)
图片数量(jpg文件个数)：4083
标注数量(xml文件个数)：4083
标注数量(txt文件个数)：4083
标注类别数：33
所在github仓库：datasets_sl
标注类别名称(注意yolo格式类别顺序不和这个对应，而以labels文件夹classes.txt为准):["barrier","cleaning-required","crack","damaged-crash-barrier","damaged-delineator","damaged-hazard-marker","damaged-kerb","delineator","deterioration","edge-drop","excessive-vegetation","face","faded-barrier-paint","faded-kerb","faded-marking","gantry-board","hazard-marker","inappropriate-diversion","informatory-board","kerb","marking","number-plate","plantation","pothole","ravelling","road","road-patch","road-stud","rumble-strips","rutting","sign-board","street-light","unauthorised-board"]
标签中文对照：护栏, 需清洁, 裂缝, 防撞护栏损坏, 轮廓标损坏, 危险警示标损坏, 路缘石损坏, 轮廓标, 破损/劣化, 边缘沉陷, 植被过度生长, 标志牌版面, 护栏漆面褪色, 路缘石褪色, 标线褪色, 龙门架标志牌, 危险警示标, 不当导流设施, 信息指示牌, 路缘石, 标线, 编号牌, 绿化带, 坑槽, 松散/剥落, 路面, 路面修补, 道钉, 减速标线, 车辙, 标志牌, 路灯, 违规标志牌
每个类别标注的框数：
barrier 框数 = 8503
cleaning-required 框数 = 61
crack 框数 = 2934
damaged-crash-barrier 框数 = 14
damaged-delineator 框数 = 3
damaged-hazard-marker 框数 = 5
damaged-kerb 框数 = 49
delineator 框数 = 13
deterioration 框数 = 7889
edge-drop 框数 = 3
excessive-vegetation 框数 = 1713
face 框数 = 9
faded-barrier-paint 框数 = 3
faded-kerb 框数 = 246
faded-marking 框数 = 2962
gantry-board 框数 = 221
hazard-marker 框数 = 287
inappropriate-diversion 框数 = 74
informatory-board 框数 = 165
kerb 框数 = 913
marking 框数 = 24834
number-plate 框数 = 131
plantation 框数 = 2165
pothole 框数 = 3
ravelling 框数 = 73
road 框数 = 4071
road-patch 框数 = 170
road-stud 框数 = 84
rumble-strips 框数 = 25
rutting 框数 = 3
sign-board 框数 = 147
street-light 框数 = 611
unauthorised-board 框数 = 1
总框数：58385
每个类别占有图片数：
barrier 占有图片数 = 4083
cleaning-required 占有图片数 = 44
crack 占有图片数 = 760
damaged-crash-barrier 占有图片数 = 14
damaged-delineator 占有图片数 = 3
damaged-hazard-marker 占有图片数 = 5
damaged-kerb 占有图片数 = 49
delineator 占有图片数 = 7
deterioration 占有图片数 = 1905
edge-drop 占有图片数 = 3
excessive-vegetation 占有图片数 = 967
face 占有图片数 = 9
faded-barrier-paint 占有图片数 = 3
faded-kerb 占有图片数 = 236
faded-marking 占有图片数 = 1441
gantry-board 占有图片数 = 214
hazard-marker 占有图片数 = 225
inappropriate-diversion 占有图片数 = 42
informatory-board 占有图片数 = 162
kerb 占有图片数 = 838
marking 占有图片数 = 4062
number-plate 占有图片数 = 121
plantation 占有图片数 = 926
pothole 占有图片数 = 3
ravelling 占有图片数 = 51
road 占有图片数 = 4065
road-patch 占有图片数 = 102
road-stud 占有图片数 = 68
rumble-strips 占有图片数 = 25
rutting 占有图片数 = 3
sign-board 占有图片数 = 138
street-light 占有图片数 = 374
unauthorised-board 占有图片数 = 1
图片分辨率：多分辨率图片,如2048x1152,2036x1146等
使用标注工具：labelImg
数据集是否增强：否
标注规则：对类别进行画矩形框
重要说明：数据集没有划分训练验证测试集需自行划分
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