注意数据集图片都是采集同一个人手势，整个数据集都是一个人模拟拍摄出来的

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)：2582

标注数量(xml文件个数)：2582

标注数量(txt文件个数)：2582

标注类别数：52

所在仓库：firc-dataset

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):["I","Play","So-so","Tomorrow","Use","Your","before","clean","drink","eat","future","good_morning","hand-gestures","happy","hate","hello","help","hour","house","how","like","live","livelong","love","more","need","new","no","past","show","sick","sleep","slow","sorry","start","stop","take care of yourself","temperature","time","today","understand","wait","wake up","want","welcome","what","where","who","why","wrong","yes","you"]

每个类别标注的框数：

I (我) 框数 = 49

Play (玩耍/播放) 框数 = 50

So-so (一般/还好) 框数 = 50

Tomorrow (明天) 框数 = 50

Use (使用) 框数 = 50

Your (你的) 框数 = 50

before (之前) 框数 = 50

clean (清洁) 框数 = 51

drink (喝) 框数 = 50

eat (吃) 框数 = 51

future (未来) 框数 = 49

good_morning (早上好) 框数 = 50

hand-gestures (手势) 框数 = 2

happy (开心) 框数 = 45

hate (讨厌) 框数 = 50

hello (你好) 框数 = 50

help (帮助) 框数 = 50

hour (小时) 框数 = 49

house (房子) 框数 = 51

how (如何) 框数 = 50

like (喜欢) 框数 = 50

live (居住) 框数 = 50

livelong (长寿) 框数 = 50

love (爱) 框数 = 49

more (更多) 框数 = 49

need (需要) 框数 = 50

new (新的) 框数 = 51

no (不) 框数 = 49

past (过去) 框数 = 48

show (展示) 框数 = 49

sick (生病) 框数 = 50

sleep (睡觉) 框数 = 45

slow (缓慢) 框数 = 1

sorry (对不起) 框数 = 49

start (开始) 框数 = 53

stop (停止) 框数 = 50

take care of yourself (照顾好自己) 框数 = 48

temperature (温度) 框数 = 116

time (时间) 框数 = 48

today (今天) 框数 = 51

understand (理解) 框数 = 49

wait (等待) 框数 = 52

wake up (醒来) 框数 = 50

want (想要) 框数 = 50

welcome (欢迎) 框数 = 50

what (什么) 框数 = 49

where (哪里) 框数 = 50

who (谁) 框数 = 50

why (为什么) 框数 = 50

wrong (错误的) 框数 = 50

yes (是的) 框数 = 100

you (你) 框数 = 50

总框数：2603

图片分辨率：640x640

使用标注工具：labelImg

标注规则：对类别进行画矩形框

重要说明：暂无

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：