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)：5092
标注数量(json文件个数)：5092
标注类别数：20
标注类别名称:["chexiang","daochazhishiqi","daokou","diangan","ganggui","guidaobiaozhizhengmian","guidaoxinhaojibeimian","guidaoxinhaojizhengmian","guidaozhedangwu","huanchongdang","huoche","jingtai daocha","liechechexiang","ren","renqun","weilan","weizhi daocha","youkai daocha","zhantai","zuokai daocha"]
对应中文类别名：["车厢", "道岔指示器", "道口", "电杆", "钢轨", "轨道标志正面", "轨道信号机背面", "轨道信号机正面", "轨道遮挡物", "缓冲挡", "货车", "静态道岔", "列车车厢", "人", "人群", "围栏", "未知道岔", "右开道岔", "站台", "左开道岔"]
每个类别标注的框数和占有图片数：
chexiang框数=109,占有图片数=80
daochazhishiqi框数=1280,占有图片数=671
daokou框数=1700,占有图片数=864
diangan框数=6429,占有图片数=1921
ganggui框数=17085,占有图片数=5091
guidaobiaozhizhengmian框数=4355,占有图片数=2179
guidaoxinhaojibeimian框数=2000,占有图片数=912
guidaoxinhaojizhengmian框数=3458,占有图片数=1613
guidaozhedangwu框数=2340,占有图片数=1051
huanchongdang框数=139,占有图片数=109
huoche框数=8,占有图片数=8
jingtai daocha框数=1480,占有图片数=844
liechechexiang框数=1192,占有图片数=828
ren框数=136,占有图片数=91
renqun框数=35,占有图片数=27
weilan框数=177,占有图片数=158
weizhi daocha框数=1508,占有图片数=925
youkai daocha框数=1313,占有图片数=943
zhantai框数=1455,占有图片数=892
zuokai daocha框数=1196,占有图片数=875
总框数：47395
使用标注工具：labelme=5.5.0
所在github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
图片预览：
标注例子：
原图（随机选16张图）：
标注绘制结果：
labelme编辑图实例：
单张图片：
